ESMRMB 2017 Congress October 19 – 21, Barcelona/ES Book of Abstracts Electronic Posters / Paper Posters / Clinical Review Posters / Software Exhibits DOI: 10.1007/s10334-017-0635-y

www.esmrmb.org

S502

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Electronic ESMRMB Posters 2017, 34th Annual Scientific Meeting, Barcelona, ES Abdominal Imaging-Clinical 527 Optimization of cervical cancer tumor volume measurement with MRI M. Shorikov1, E. Tarachkova2, I. Gubskiy3, V. Panov2, N. Schimanowsky4 1 Radiology, Russian Cancer Research Center, Moscow/RUSSIAN FEDERATION, 2Radiology, Russian Medical Academy of Continuous Professional Education, Moscow/RUSSIAN FEDERATION, 3 Radiology, Russian National Research Medical University named after N.I. Pirogov, Moscow/RUSSIAN FEDERATION, 4 Pharmacology and radiobiology, Russian National Research Medical University named after N.I. Pirogov, Moscow/RUSSIAN FEDERATION Purpose/Introduction: There’re different ways to measure the tumor volume [1–5] which is very important for subsequent radiotherapy.The purpose was to optimize the measurement of the tumor volume before radiotherapy. Subjects and Methods: 22 patients with histologically verified cervical cancer (CC) were enrolled in this study. Axial DWI and T2WI with 3 mm thick slices were used, the tumor volume was estimated as the sum of the tumor areas on each slice multiplied by thickness. For comparison it was also measured using only each 2nd, 3rd, 4th or 5th slice or with ellipse, cylinder and cone estimations using: (1) three axes or (2) maximum area and two axes. Results: T2w and DWI initial measurements didn’t differ significantly (p [ 0.2). Ellipse, cylinder and cone estimations were significantly different to each slice measurement (30–100% difference, p \ 0.05). On T2WI using only each 2 nd slice is not significantly different to using each one (p [ 0.05) but the usage of each 3rd slice is (p \ 0.05). On DWI the usage of each 3rd slice is not different from the initial measurement (p [ 0.8) but the usage of each 4th slice is (p \ 0.05). Box and whisker plot for different measurement methods rate if error on DWI is presented on Fig. 1.

123

Discussion/Conclusion: The most precise and the least time-consuming method to estimate the tumor volume is on DWI with the measurement of the tumor areas on each 3rd slice multiplied by distance between used slices (slice thickness = 3 mm). References: 1. Hagen C. P., Mouritsen A., Mieritz M. G., Tinggaard J., WohlfahrtVeje C., Fallentin E., Brocks V., Sundberg K., Jensen L. N., Juul A., Main K. M.: Uterine volume and endometrial thickness in healthy girls evaluated by ultrasound (3-dimensional) and magnetic resonance imaging. Fertil Steril 2015, 104(2):452–459.e452. 2. Karpagam S., Gowri S.: Brain Tumor Growth and Volume Detection by Ellipsoid-Diameter Technique Using MRI Data. International Journal of Computer Science 2012, 9(2):121–126. 3. Montelius M., Ljungberg M., Horn M., Forssell-Aronsson E.: Tumour size measurement in a mouse model using high resolution MRI. BMC Med Imaging 2012, 12:12. 4. Olver F. W. J., Lozier D. W., Boisvert R. F., Clark C. W.: NIST Handbook of Mathematical Functions: Cambridge University Press; 2010. 5. Spithoven E. M., van Gastel M. D., Messchendorp A. L., Casteleijn N. F., Drenth J. P., Gaillard C. A., de Fijter J. W., Meijer E., Peters D. J., Kappert P., Renken R. J., Visser F. W., Wetzels J. F., Zietse R., Gansevoort R. T.: Estimation of Total Kidney Volume in Autosomal Dominant Polycystic Kidney Disease. Am J Kidney Dis 2015.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

528 Target fusion-biopsy in prostate cancer diagnostics F. Kossov1, V. Panov2, B. Kamolov3, I. Abdullin3, E. Baranova3, I. Tyurin2 1 Radiology, RONC for name Blohina, Moscow/RUSSIAN FEDERATION, 2Radiology, Russian Medical Academy of Continuous Professional Education, Moscow/RUSSIAN FEDERATION, 3 Urology, European Medical Centre, Moscow/RUSSIAN FEDERATION Purpose/Introduction: Nowadays standard prostate biopsy (SB) is an integral part of prostate cancer (PCa) diagnostics. PCa detection rate using SB varies from 31 to 42%, however the number of falsenegative results remains much too high. Fusion of mpMRI data with US images in US-guided target biopsy or target fusion-biopsy (TB) (hybrid technology) into routine clinical practice would significantly improve the detection of medium-risk cancer (MRC) and high-risk cancer (HRC) both in initial biopsy (IB) and in patients with negative biopsy results. Subjects and Methods: 47 patients (men at the age of 44 up to 71 years (median-62) with suspected PCa were examined. The criteria of patients inclusion in examination was the presence of: (1) PSA [ 4 ng/ml and/or positive/controversal results of the digital rectal examination; (2) mpMRI data (PI-RADS [ 2). The examination was held in two steps: (1) mpMRI of prostate; (2) TB. Results: PCa was detection in 32 primary cases and 3 cases with previously negative SB. The benign prostatic hyperplasia was founded in 12 cases. TB statistically significant (Spearman correlation coefficient—0.8, p \ 0.0001) is highly competitive in high malignancy potential nodes detection with SB, however according to regressive analysis report the HRC detection accuracy rate using TB appeared higher in comparison with SB (p \ 0.0001). MRI data showed statistically significant correlation with the TB results (p \ 0.02, TB correlation coefficient 0.7). Discussion/Conclusion: The TB technology implementation provides an opportunity to decrease the number of punctures made for detection of the histological material with the highest malignancy rate. TB technology implementation exposed non-sensibility to non-clinically significant PC. The results of the recent advances in diagnostic imaging in conjunction with hybrid technology leave no doubts that the further prostate biopsy development vector is to be narrowed to the ‘‘targeted and seeing’’ biopsy conception, that in turn may significantly increase the detection of MRC and HRC. The implementation of MRI/US Fusion biopsy technology lets to significantly increase the detection of clinically relevant forms of PCa, correctly detect the clinical state of disease, accurately stratify the risks. All these factors are highly important for the further therapeutic decisions for each patient. References: 1. Scattoni V, Maccagnano C, Capitanio U, b lp. Random biopsy: When, how many and where to take the cores? World J Urol 2014;32:859–69. doi:10.1007/s00345-014-1335-0. 2. Presti JC. Prostate biopsy: How many cores are enough? B: Urologic Oncology: Seminars and Original Investigations. 2003. 135–40. doi:10.1016/S1078-1439(03)00006-1. 3. Weiner AB, Patel SG, Etzioni R, b lp. National trends in the management of low and intermediate risk prostate cancer in the United States. J Urol 2015;193:95–102. doi: 10.1016/j.juro.2014.07.111. 4. Bjurlin MA, Wysock JS, Taneja SS. Optimization of prostate biopsy: Review of technique and complications. Urol. Clin. North Am. 2014;41:299–313. doi:10.1016/j.ucl.2014.01.011.

S503 5. Djavan B, Milani S, Remzi M. Prostate biopsy: who, how and when. An update. Can. J. Urol. 2005;12 Suppl 1. 6. Rodrı´guez-Covarrubias F, Gonza´lez-Ramı´rez A, Aguilar-Davidov B, b lp. Extended sampling at first biopsy improves cancer detection rate: Results of a prospective, randomized trial comparing 12 versus 18-core prostate biopsy. J Urol 2011;185:2132–6. doi: 10.1016/j.juro.2011.02.010. 7. Ploussard G, Nicolaiew N, Marchand C, b lp. Prospective evaluation of an extended 21-core biopsy scheme as initial prostate cancer diagnostic strategy. Eur Urol 2014;65:154–61. doi: 10.1016/j.eururo.2012.05.049. 8. Pokorny MR, De Rooij M, Duncan E, b lp. Prospective study of diagnostic accuracy comparing prostate cancer detection by transrectal ultrasound-guided biopsy versus magnetic resonance (MR) imaging with subsequent mr-guided biopsy in men without previous prostate biopsies. Eur Urol 2014;66:22–9. doi: 10.1016/j.eururo.2014.03.002. 9. Sonn GA, Chang E, Natarajan S, b lp. Value of targeted prostate biopsy using magnetic resonance-ultrasound fusion in men with prior negative biopsy and elevated prostate-specific antigen. Eur Urol 2014;65:809–15. doi:10.1016/j.eururo.2013.03.025. 10. Barentsz JO, Richenberg J, Clements R, b lp. ESUR prostate MR guidelines 2012. Eur Radiol 2012;22:746–57. doi: 10.1007/s00330-011-2377-y. 11. Somford DM, Hamoen EH, F€ Utterer JJ, b lp. The Predictive Value of Endorectal 3 Tesla Multiparametric Magnetic Resonance Imaging for Extraprostatic Extension in Patients with Low, Intermediate and High Risk Prostate Cancer. J Urol 2013;190:1728–34. doi: 10.1016/j.juro.2013.05.021. 12. Delongchamps NB, Zerbib M. Role of magnetic resonance imaging before initial biopsy: Comparison of magnetic resonance imaging-targeted and systematic biopsy for significant prostate cancer detection. Eur. Urol. 2012;61:622–3. doi:10.1016/j.eururo.2011.12.038. 13. Hoeks CMA, Schouten MG, Bomers JGR, b lp. Three-tesla magnetic resonance-guided prostate biopsy in men with increased prostate-specific antigen and repeated, negative, random, systematic, transrectal ultrasound biopsies: Detection of clinically significant prostate cancers. Eur Urol 2012;62:902–9. doi: 10.1016/j.eururo.2012.01.047. 14. Rosenkrantz AB, Verma S, Choyke P, b lp. Prostate Magnetic Resonance Imaging and Magnetic Resonance Imaging Targeted Biopsy in Patients with a Prior Negative Biopsy: A Consensus Statement by AUA and SAR. J Urol 2016;196:1613–8. doi: 10.1016/j.juro.2016.06.079. 15. Cash H, Maxeiner A, Stephan C, b lp. The detection of significant prostate cancer is correlated with the Prostate Imaging Reporting and Data System (PI-RADS) in MRI/transrectal ultrasound fusion biopsy. World J Urol 2016;34:525–32. doi:10.1007/s00345-015-1671-8.

529 Qualitative and quantitative evaluation of image quality of sagittal and axial T2-Weighted Fast Spin-Echo MRI examinations of the Female Pelvis S. Al Dahery, L. Rainford, A. McGee Health Science, UCD, Dublin/IRELAND Purpose/Introduction: To evaluate the resultant image quality of female pelvis scans across clinical centres following axial and sagittal T2-weighted imaging (T2 W) techniques at 1.5T, while focusing on the normal uterine body anatomy.

123

S504 Subjects and Methods: Retrospective axial and sagittal T2 W images from centres across Ireland (n = 10) and the Kingdom of Saudi Arabia (n = 9) for female pelvic MR scans performed on 1.5T MRI scanners (n = 117) were analysed. Quantitative findings for SNR and CNR were obtained by applying ROIs on normal myometrium-tissue, gluteus maximus muscle and background noise. Data sets were then categorised into two groups defined by sequence parameters, namely by slice thickness applied. Images were evaluated qualitatively for presence of motion artefacts, visualisation of the pelvic anatomical structures including the myometrium, endometrium and junctional zone for T2 W images (axial and sagittal). Uterine MR image quality; SNR, CNR, and spatial resolution and overall image quality were evaluated. Inter-observer variability was determined by weighted kappa statistics in ‘‘R program’’ to measure the degree of agreement between readers (n = 8). Results: Median Irish SNR values were higher for all sequences compared to the Saudi findings. A significant difference in SNR between the two countries for T2 W axial images was noted upon Bonferroni correction. No significant differences were noted for SNR when T2 W sagittal images were compared geographically. CNR findings returned no significant differences, for both sequences. Qualitative analysis identified differences in image quality criteria scores for motion artefact, junctional zone, and endometrium visualization for T2 W axial images from protocol 1 using slice thickness of 4 mm and below; compared to protocol 2 using slice thickness of 5 mm and above (p \ 0.05). Overall image quality demonstrated statistical significant difference between protocols. The presence of motion artefact and readings for SNR and spatial resolution were significantly different between protocol 1 (slice thickness = 3.5 mm) and protocol 2 (slice thickness = 5 mm) for T2 W-2D-FSE-SAG with the readers preferring protocol 1 for scanning the T2 W sagittal images. Differences in quantitative and qualitative findings were noted. Discussion/Conclusion: MR sequence parameters have the potential to affect the visualization of uterine anatomical structures and overall image quality. In addition to slice thickness, sequence parameters such as echo time (TE), repetition time (TR), and the flip angle; can affect the SNR and image quality. An appropriate knowledge of NMR physics is required to fully understand the factors affecting MR image quality and both quantitative and qualitative review of MRI images is required when evaluating scan images. References: Beddy, P., Rangarajan, R. D., Kataoka, M., Moyle, P., Graves, M. J. & Sala, E. (2011). T1-weighted fat-suppressed imaging of the pelvis with a dual-echo Dixon technique: initial clinical experience. Radiology, 258, 583–589. Froehlich, J. M., Metens, T., Chilla, B., Hauser, N., Klarhoefer, M. & Kubik-Huch, R. A. (2012). Should less motion sensitive T2-weighted BLADE TSE replace Cartesian TSE for female pelvic MRI? Insights into imaging, 3, 611–618. Gadermann, A. M., Guhn, M. & Zumbo, B. D. (2012). Estimating ordinal reliability for Likert-type and ordinal item response data: A conceptual, empirical, and practical guide. Practical Assessment, Research & Evaluation, 17, 1–13. Kataoka, M., Kido, A., Koyama, T., Isoda, H., Umeoka, S., Tamai, K., Nakamoto, Y., Maetani, Y., Morisawa, N. & Saga, T. (2007). MRI of the female pelvis at 3T compared to 1.5 T: Evaluation on highresolution T2-weighted and HASTE images. Journal of Magnetic Resonance Imaging, 25, 527–534. Lane, B. F., Vandermeer, F. Q., Oz, R. C., Irwin, E. W., Mcmillan, A. B. & Wong-You-Cheong, J. J. (2011). Comparison of sagittal T2weighted BLADE and fast spin-echo MRI of the female pelvis for motion artifact and lesion detection. American Journal of Roentgenology, 197, W307–W313.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

530 MRI as a triage test in prostate cancer diagnostic algorithm in biopsy-naı¨ve men: prospective study A. Pavlicˇko, J. Votrubova´ Radiology, Thomayer hospital, Prague/CZECH REPUBLIC Purpose/Introduction: Current diagnostic algorithm has its weakness in a substantial number of unnecessary biopsies and detection of clinically insignificant cancer. The aim of study is to show potential benefits of MRI as a triage test in a diagnostic algorithm of prostate cancer. Subjects and Methods: Inclusion criteria: (1) age B75 years, (2) elevated PSA above age-specific limit or suspected PSA levels or its derivatives according to current guidelines, (3) negative digital rectal examination, Exclusion criteria: (1) previous TRUS-biopsy, (2) prostate surgery, (3) impossibility to undergo MRI, 4.contra-indications for biopsy. Multiparametric MRI and MRI/TRUS fusion guided targeted biopsy with systematic biopsy (12 cores) performed in 132 patients. Multiparametric MRI performed on a Signa 1.5T GE HDXT using surface and endorectal coil. Mean age of patients: 62 years. Mean and median PSA: 7.15 ng/ml and 6.7 ng/ml resp. Mean prostate volume and volume of transitional zone: 58.5 ml and 31.9 ml resp. Overall detection rates and detection rates of clinically significant prostate cancer (CSPCA) were stratified according to PIRADS score version 1 into 5 groups: MRI negative (PIRADS \ 3), PIRADS 3, 4, 5 and NA (MRI positive, but precise PIRADS score was not available). Number of avoided biopsies and number of missed prostate cancers were counted for significant and insignificant prostate cancer. CSPC was defined as the presence of cancer with subsequent characteristics: Gleason score [6 or 3 and more positive biopsy cores or [50% of the presence of prostate cancer in a biopsy core. Results: of detection rates: PIRADS Number of patients Overall prostate cancer Clinically significant PC Insignificant PC

<3 3 4 5 NA 29 27 55 15 6 2(6.9%) 11(40.7%) 36(65.5%) 14(93.3%) 2(33%)

Σ 132 65(49.2%)

1 (3.4%) 3 (11.1%) 30 (54.5%) 13 (86.7%) 2 (33%) 49 (37.1%) 1 (3.4%)

8 (29.6%) 6 (10.9%) 1 (6.7%)

0 (0.0%) 16 (12.1%)

If cut-off level was PIRADS \3, 21.9% of biopsies could be avoided, 6.3% of clinically insignificant prostate cancer would not be diagnosed but 2.0% of clinically significant prostate cancer would be missed. If cut-off level was PIRADS B3, 42.4% of biopsies could be avoided, 56.3% of clinically insignificant prostate cancer would not be diagnosed but 8.1% of clinically significant prostate cancer would be missed. Discussion/Conclusion: If MRI served as triage test in diagnostic algorithm of prostate cancer, significant number of biopsies could be avoided and considerable amount of clinically insignificant prostate cancer would not be diagnosed. On the other hand, a small but not negligible number of clinically significant prostate cancer would be missed. References: Supported by the Agency for Healthcare Research MZCˇR, project number 15-27047A.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S505

531 Local recurrence of prostate carcinoma after ExternalBeam Radiotherapy (EBRT): comparative sensitivities of multiparametric MRI Sequences P.P. Arcuri1, S. Roccia2, D. Lagana`3, D. Pingitore4, E. Mazzei4, G. Fodero1 1 Radiology, A.O. Pugliese-De Lellis, Catanzaro/ITALY, 2Centro Medicina Legale, INPS, Catanzaro/ITALY, 3Radiology, Universita` Magna Graecia, Catanzaro/ITALY, 4radioterapy, A.O. Pugliese-De Lellis, Catanzaro/ITALY Purpose/Introduction: The aim of our study was to compare the respective accuracies of T2 sequence, diffusion-weighted imaging (DWI), dynamic contrast-enhanced (DCE) and 3D 1H–MR spectroscopy (MRS) to detect local prostate cancer recurrence, after EBRT. Subjects and Methods: This retrospective study included 19 men (median age, 69 years) affected by recurrence of prostate carcinoma post-EBRT (median dose, 7800 cGy), ascertained on the basis of transrectal US-guided biopsy performed after a suspicion induced by a prostate-specific antigen (PSA) level above 0.2 ng/mL or more. MR images were acquired before treatment and 3 months after treatment at least. A pelvic phased-array coil with 32 channels was used. During the acquisition, the bladder was in the state of mild repletion. The study protocol included: Diffusion-weight imaging (DWI) with b-values of 0, 1000, 1500 and 2000 s/mm2, dynamic contrast-enhanced imaging (DCE) and 3D 1H–MR spectroscopy (MRS). The Apparent Diffusion Coefficient (ADC), the volume transfer constant (K trans) and the rate constant (Kep) were calculated. On ADC map a nodule was considered positive for malignancy if the cut-off value was equal to or lower than 1.000 9 10-3 mm2/s. For spectroscopic analysis, a voxel was considered positive for malignancy for a ratio value threshold up to 0.86 or when only a choline peak was detected. In DCE were considered: a qualitative analysis of time-signal intensity curves such as high maximum relative enhancement and short time to peak, high wash-in rate and wash-out slope of the curve, to diagnose malignant tissue. The Fisher test was used to determine significant differences in sensitivity of multiparametric MRI. A p value less than 0.05 was considered as indicative of statistically significant differences. Two readers independently evaluated the results. Results: Local recurrence was found more often in the peripheral zone (74%) than in the transitional zone (26%). Sensitivity with T2weighted sequence was 78%, with ADC map was 81%.

The mean ADC value in malignancy areas was 0.73 9 10-3 mm2/s. DCE sequence (Fig. 2,), was positive in 13 cases (69%) Fig. 3.

123

S506

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 6) Lopes Dias J, Lucas R, Magalha˜es Pina J, Joa˜o R, Costa NV, Leal C, Bilhim T, Campos Pinheiro L, Mateus Marques R. Post-treated prostate cancer: normal findings and signs of local relapse on multiparametric magnetic resonance imaging. Abdom Imaging. 2015 Oct;40(7):2814–38. 7) Oguz Akin, David Gultekin, Hebert Alberto Vargas, Junting Zheng, Chaya Moskowitz, Xin Pei, Dahlia Sperling, Lawrence Schwartz, Hedvig Hricak, and Michael Zelefsky. Incremental Value of Diffusion Weighted and Dynamic Contrast Enhanced MRI in the Detection of Locally Recurrent Prostate Cancer after Radiation Treatment: Preliminary Results. Eur Radiol. 2011 Sep; 21(9): 1970–1978. 8) Panebianco V, Barchetti F, Grompone MD, Colarieti A, Salvo V, Cardone G, Catalano C. Magnetic resonance imaging for localization of prostate cancer in the setting of biochemical recurrence. Urol Oncol. 2016 Jul;34(7):303–10. 9) Roy C et al. Comparative sensitivities of functional MRI sequences in detection of local recurrence of prostate carcinoma after radical prostatectomy or external-beam radiotherapy. AJR Am J Roentgenol. (2013).

532 Comparison of Reduced field-of-View DiffusionWeighted Magnetic Resonance Imaging with Conventional Diffusion-Weighted Imaging in cervical cancer

For 3D 1H-MRS sensitivity was 64% (there was found a significant choline peak with a ratio over 0.8 in 12 patients). Discussion/Conclusion: The highest sensitivity was obtained with DWI sequence. T2-weighted imaging remains essential for the morphologic analysis. DCE and even more spectroscopy showed low sensitivity probably correlated with vascularization and metabolites modifications after EBRT. References: 1) Abd-Alazeez M, Ramachandran N, Dikaios N, Ahmed HU, Emberton M, Kirkham A, Arya M, Taylor S, Halligan S, Punwani S. Multiparametric MRI for detection of radiorecurrent prostate cancer: added value of apparent diffusion coefficient maps and dynamic contrast-enhanced images. Prostate Cancer Prostatic Dis. 2015 Jun;18(2):128–36. 2) Ahmed HU, Cathcart P, McCartan N, Kirkham A, Allen C, Freeman A, Emberton M. Focal salvage therapy for localized prostate cancer recurrence after external beam radiotherapy: a pilot study. Cancer. 2012 Sep 1;118(17):4148–55. 3) Alonzo F, Melodelima C, Bratan F, Vitry T, Crouzet S, Gelet A, Rouvie`re O. Detection of locally radio-recurrent prostate cancer at multiparametric MRI: Can dynamic contrast-enhanced imaging be omitted? Diagn Interv Imaging. 2016 Apr;97(4):433–41. 4) Daniel Corey Oppenheimer, M.D., Eric P Weinberg, Gary M Hollenberg, and Steven P Meyers. Multiparametric Magnetic Resonance Imaging of Recurrent Prostate Cancer. J Clin Imaging Sci. 2016; 6:18. Published online 2016 Apr 29. 5) Donati OF, Jung SI, Vargas HA, Gultekin DH, Zheng J, Moskowitz CS, Hricak H, Zelefsky MJ, Akin O. Multiparametric prostate MR imaging with T2-weighted, diffusion-weighted, and dynamic contrast-enhanced sequences: are all pulse sequences necessary to detect locally recurrent prostate cancer after radiation therapy? Radiology. 2013 Aug;268(2):440–50.

123

J. Hwang, S.S. Hong, H.-J. Kim, Y.-W. Chang Department of Radiology, Soonchunhyang University College of Medicine, Seoul Hospital, Seoul/KOREA, REPUBLIC OF Purpose/Introduction: To determine qualitative and quantitative value of reduced-field-of view (FOV) diffusion-weighted imaging (DWI) in patient with cervical cancer in comparison with conventional DWI. Subjects and Methods: This study included 22 patients with pathologically confirmed cervical cancer who underwent MR examination before therapy. Two observers independently reviewed reduced FOV DWI and conventional DWI. The study variables in qualitative analysis included anatomic detail, lesion conspicuity, artifacts and overall image quality. Quantitative analysis was performed by measuring ADC value of the tumor. Wilcoxon signed-rank test was used to compare qualitative scores and the mean ADC value between two DWI sequences. Results: Reduced FOV DWI achieved significantly better anatomic detail (3.75 ± 0.44 at b = 0 s/mm2 and 3.41 ± 0.54 at b = 1000 s/ mm2), lesion conspicuity (3.53 ± 0.57 at b = 0 s/mm2 and 3.60 ± 0.50 at b = 1000 s/mm2), artifact (3.11 ± 0.58 at b = 0 s/ mm2 and 2.91 ± 0.64 at b = 1000 s/mm2), and overall image quality (3.34 ± 0.61 at b = 0 s/mm2 and 3.29 ± 0.55 at b = 1000 s/mm2) compared to conventional FOW DWI (anatomic detail, 3.0 ± 0.37 at b = 0 s/mm2 and 2.93 ± 0.50 at b = 1000 s/mm2; lesion conspicuity, 2.77 ± 0.43 b = 0 s/mm2 and 3.07 ± 0.52 at b = 1000 s/mm2; artifact, 2.93 ± 0.50 b = 0 s/mm2 and 2.70 ± 0.59 at b = 1000 s/ mm2; overall image quality 2.95 ± 0.43 at b = 0 s/mm2 and 2.82 ± 0.49 at b = 1000 s/mm2) (p \ 0.05). There was no significant difference in the mean tumor ADC value between two DWI sequences (0.990 9 10-3 mm2/s ± 0.364 at reduced FOV DWI vs. 1.253 9 10-3 mm2/s ± 0.387 at conventional DWI) (p = 0.067). Discussion/Conclusion: Reduced FOV DWI yielded better image quality including anatomic detail and lesion conspicuity with decreased artifact.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 References: 1. Feng Z. et al. Comparison of field-of-view (FOV) optimized and constrained undistorted single shot (FOCUS) with conventional DWI for the evaluation of prostate cancer Clinical Imaging 2015;39: 851–855. 2. Kim H. et al. Reduced Field-of-View Diffusion-Weighted Magnetic Resonance Imaging of the Pancreas: Comparison with Conventional Single-Shot Echo-Planar Imaging Korean J Radiol 2015;16(6):1216–1225. 3. Chao Ma et al. High resolution diffusion weighted magnetic resonance imaging of the pancreas using reduced field of view singleshot echo-planar imaging at 3 T Magnetic Resonance Imaging 32 (2014) 125–131.

533 Magnetic resonance imaging findings of atypical uterine leiomyomas A. Er1, G. Pekindil2, M. Gok3, S. Guneyli4, A.R. Kandiloglu5, A. Goker6 1 RADIOLOGY, VAN RESEARCH HOSPI˙TAL, VAN/TURKEY, 2 RADIOLOGY, CELAL BAYAR UNIVERSITY, MANISA/TURKEY, 3 RADIOLOGY, ADNAN MENDERS UNIVERSITY, AYDIN/TURKEY, 4 RADIOLOGY, BULENT ECEVIT UNIVERSITY, ZONGULDAK/ TURKEY, 5PATHOLOGY, CELAL BAYAR UNIVERSITY, MANISA/ TURKEY, 6OBSTETRICS, CELAL BAYAR UNIVERSITY, MANISA/ TURKEY Purpose/Introduction: Uterine leiomyomas are the most common woman pelvic tumors, and it may be difficult to distinguish the atypical ones from the leiomyosarcomas. We aimed to evaluate magnetic resonance imaging (MRI) findings of uterine leiomyomas and distinguish different histopathological types of leiomyomas by evaluating diffusion weighted imaging (DWI) and apparent diffusion coefficient (ADC) values. Subjects and Methods: Thirty-eight female patients aged 27–66 (median 43) years who underwent MRI between July 2011 and January 2013 were included in this retrospective study. The size, location, signal intensity of the tumor on T1-, T2-, diffusion-weighted images, and ADC values were evaluated. The histopathological types of leiomyomas were compared according to the signal intensity on DWI and ADC values. Mann–Whitney U, v2, and Kruskal–Wallis tests were used for statistical analyses. Results: A total of 69 leiomyomas (54 typical, 5 degenerated, and 10 cellular) were demonstrated. The mean ± standard deviation (SD) of leiomyoma size was 5.37 ± 3.40 cm. The locations were as follows; 53 intramural, 12 subserosal and 4 submucosal. On DWI, all of the cellular leiomyomas, 20% of degenerated leiomyomas, and 5.6% of typical leiomyomas were hyperintense (P \ 0.001). The mean ± SD value of b-800 ADC value (10-3 mm2/ s) of the cellular leiomyomas, typical leiomyomas, and degenerated leiomyomas were 1.11 ± 0.12, 1.29 ± 0.19, and 1.86 ± 0.17, respectively (P = 0.01). Additionally, a difference of ADC values between the leiomyomas greater than 8 cm and smaller than 8 cm was found (P \ 0.05). Discussion/Conclusion: DWI and ADC values can be effectively used to distinguish atypical uterine leiomyomas from the typical ones. The diagnosis of an atypical leiomyoma should be considered for the lesions suspicious for leiomyosarcoma. References: 1. Tamai K, Koyama T.,Saga T, Morisawa N., Mikami Y., Togashi K., The utility of diffusion weighted MR imaging for differentiating

S507 uterine sarcomas from benign leiomyomas, Eur Radiol 2008, 18: 723–730. 2. Bammer, R., Liu, C., Po, J., Moseley, M.E. Diffusion weighted magnetic resonance imaging. In: Edelman, Hesselink, Zlatkin, Crues. Clinical magnetic resonance imaging, 3rd ed, Philadelphia, Saunders Elsevier, 2006; 288–319.

534 Prostate perfusion study: new techniques for acceleration and motion control D. Kupriyanov1, L. Araslanova2, E. Demchenko2 1 Imaging Systems, Philips Healthcare, Moscow/RUSSIAN FEDERATION, 2Radiology department, : Rostov-on-Don Regional Diagnostic Center, Rostov-on-Don/RUSSIAN FEDERATION Purpose/Introduction: To improve diagnostic accuracy of high resolution prostate DCE study, new acceleration techniques could be used: k-space viewsharing (4dTRAK) [1] and time domain and k-space undarsampling (k-t SENSE) [2]. Subjects and Methods: The study was performed for 15 volunteers, pre-planned for routine prostate DCE study. Contrast injection was splitted into two pases: both standard and accelerated DCE scans were aquired with 5 min delay. Image SNR, Contrast Enhancement Ratios and Artifacts scoring were done for left central, left periferal, right central and right periferal parts of prostate. Dynamic perfusion curves were estimated for both sequences. Results: Compared with standard perfusion MR, image quality is improved and artifacts are reduced for K-t Sense technique. Perfusion curve demonstrates much lower level of motion induced artifacts (Pic1.). At the same time, Contrast Enhancement Ratio is also decreased for k-t Sense perfusion scans.

Discussion/Conclusion: k-t SENSE accelerated perfusion technique demonstrates acceptable diagnostic accuracy for prostate studies. Compared with classic perfusion MR sequencies, image quality is improved and motion induced artifacts are reduced. References: [1] - Hadizadeh DR, Gieseke J, Beck G, Geerts L, Kukuk GM, Bostro¨m A, Urbach H, Schild HH, Willinek WA, View-sharing in keyhole imaging: Partially compressed central k-space acquisition in time-resolved MRA at 3.0 T. 2011 Nov;80(2):400–6. [2] - Jeffrey Tsao, Peter Boesiger, and Klaas P. Pruessmann. k-t BLAST and k-t SENSE: Dynamic MRI With High Frame Rate Exploiting Spatiotemporal Correlations. Magnetic Resonance in Medicine 50:1031–1042 (2003).

123

S508

535 Evaluation of diffusion-weighted Imaging in the context of multi-parametric MRI of the prostate in the assessment of suspected low volume prostatic carcinoma I. Papadopoulos1, J. Phillips2, R. Evans2, N. Fenn3, S. Shermer4 1 Physics & Medicine, Swansea University, Swansea/UNITED KINGDOM, 2Medical School, Swansea University, Swansea/UNITED KINGDOM, 3Urology, Morriston Hospital, Swansea/UNITED KINGDOM, 4College of Science, Dept of Physics, Swansea University, Swansea/UNITED KINGDOM Purpose/Introduction: multi-parameteric MRI (mp-MRI) significantly increase the reliability of MRI-based detection of peripheral zone prostate cancer [1]. Data from a mp-MRI study for possible early-stage prostate cancer was assessed with a view to creating an efficient clinical screening protocol. Subjects and Methods: 46 patients between 48 and 85 years, from which 20 have probably malignant prostatic areas based on three criteria, had mp-MRI examination using a 3T scanner. Data of diffusion-weighted and T2-weighted imaging were analysed to evaluate the combination of the two techniques as a potential diagnostic tool of prostate cancer detection, staging and guided biopsies. Also, data of MR spectroscopy and dynamic contrast enhanced imaging were acquired to assess the correlation of PIRADS [2] scores for these MR techniques and overall PIRADS scores. Results: Quantification of the noise floor in the DWI images and careful fitting of the data to avoid spurious kurtosis results due to noise floor contamination suggest that the mono-exponential model provides a very good fit to the data and there is no evidence of nonGuassian diffusion for b-values up to 1000 s/mm2. This precludes the use of kurtosis as a biomarker for prostate cancer in our case. The ADC scores for healthy and probably malignant areas are significantly lower for the latter in all but one patient suggesting it a reliable biomarker. Discussion/Conclusion: The results suggest that a simplified mpMRI protocol combining T2w and DWI may be a good compromise for a cost- and time efficient, early-stage prostate cancer screening programme. Combining robust MR biomarkers for prostate cancer can be reliably quantified and appear well suited for general clinical practice [3]. References: [1] Delongchamps, Nicolas Barry, Mathieu Rouanne, Thierry Flam, Fre´de´ric Beuvon, Mathieu Liberatore, Marc Zerbib, and Franc¸ois Cornud. ‘‘Multiparametric magnetic resonance imaging for the detection and localization of prostate cancer: combination of T2weighted, dynamic contrast-enhanced and diffusion-weighted imaging.’’ BJU international 107, no. 9 (2011): 1411–1418. [2] Barentsz, Jelle O., et al. ‘‘ESUR prostate MR guidelines 2012.’’ European radiology 22.4 (2012): 746–757. [3] Ioannis Papadopoulos, Jonathan Phillips, Rhodri Evans, Neil Fenn, Sophie Shermer. ‘‘Evaluation of Diffusion Weighted Imaging in the Context of Multi-Parametric MRI of the Prostate in the assessment of suspected low volume prostatic carcinoma’’—Submitted to Magnetic Resonance Imaging.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

536 Advantages of MR-enterography for definition of the activity in Crohn’s disease O. Subbotina, M. Rezakova, A. Shevshenko Department of MRI, State Research Institute of Physiology and Fundamental Medicine SB RAMS, Novosibirsk/RUSSIAN FEDERATION Purpose/Introduction: The attention of scientists of all countries to Crohn’s disease does not weaken. Crohn’s Disease is principally characterized by chronic and recurrent inflammation of the gastrointestinal tract, most commonly found in the ileo-colonic region. Only a combination of certain symptoms makes it possible to establish a diagnosis of Crohn’s disease and to estimate the prognosis of the disease, therefore, various variants of prognostic assessment scales are offered, taking into account the form and severity of the disease, laboratory and instrumental studies that allow the doctor to better orientate in choosing the specific type of treatment. Therefore, the purpose of this paper is to evaluate the capabilities of using MRenterography and endoscopy with biopsy in Crohn’s Disease. Subjects and Methods: In this paper the results of endoscopy and MR-enterography (MRE) were analyzed. A total of 50 people with Crohn’s disease, men and women, aged 19–62 years. The inclusion of patients into the study was carried out in the presence of colonoscopy with examination of the terminal ileum and biopsy and MRE (within 5–7 days of one another). The control group consisted of 30 cases without inflammatory bowel disease. The study was conducted on the base of the MRI system General Electric Discovery MR750 W (3.0 T) using a standard protocol for the examination of the bowel. The following parameters assessed on MRE: bowel wall thickness, mesenteric edema, fibrofatty proliferation, stricturing, obstruction, abscesses, and fistulae. Subsequently, we compared the results of endoscopy and MRE.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results: We did not observe any significant differences in the definition of bowel inflammatory activity methods endoscopy and MRE. In this way MRE is not only accurate for assessing mucosal finding, but also extraintestinal changes of small and large intestine. Discussion/Conclusion: Certainly endoscopy with biopsy is the ‘‘gold standard’’ of diagnosis, evaluate the degree of bowel inflammatory activity in Crohn’s Disease. MRE is noninvasive objective assessment of disease activity, bowel damage and extraintestinal changes. In this regard it is established as the first-line imaging test for patients with Crohn’s disease for detection, dynamic monitoring. References: Stidham RW, Cross RKWesterland O., Griffin N. Magnetic Resonance Enterography in Crohn’s Disease. Semin Ultrasound CT MR 2016; 37(4):282–291. Pariente B., et al. Development of the Crohn’s disease damage score, the Lemann score. Inflamm Bowel Dis 2011; 17 (6): 1415–1422. Stidham RW, Cross RK. Endoscopy and cross-sectional imaging for assessing Crohn’s disease activity. Tech Gastrointest Endosc. 2016 Jul;18(3):123–130.

537 Early Stage Liver Fibrosis: Parenchymal biopsy versus ADC measurements O. Tuncyurek1, F. Duzgun2, F. Turkdogan1, E. Ertekin1, A. Kandemir3, V.S. Ozturk1, I.K. Omurlu4, N. Culhaci5, G. Pekindil2, Y. Ozsunar1 1 Radiology, Adnan Menderes University School of Medicine, AYDIN/ TURKEY, 2Radiology, CELAL BAYAR UNIVERSITY SCHOOL OF MEDICINE, MANISA/TURKEY, 3GASTROENTEROLOGY, Adnan Menderes University School of Medicine, AYDIN/TURKEY, 4 BIOSTATISTICS, Adnan Menderes University School of Medicine, AYDIN/TURKEY, 5PATHOLOGY, Adnan Menderes University School of Medicine, AYDIN/TURKEY Purpose/Introduction: There are many parenchyma biopsy scoring systems for liver disease, and the scores may vary among observers. The aim the study to reveal the role of DWI technique in the differentiation of early-stage fibrosis and to discuss the importance of imaging as alternative to biopsy. Subjects and Methods: Fifty-eight diagnosed patients were evaluated with DWI during their clinical follow-up. Biopsies were scored for fibrosis by ISHAK scores. The differences in ADC measurements among the cases grouped as F0 (n = 20), F1-3 (n = 10), and F4-6 (n = 28) were compared. Both centers made the evaluations without being aware of the observer and the histopathological scores of the patients. The ADC values were statistically evaluated with the ISHAK score. Results: The ADC values of F1-3 and F4-6 were significantly different from that of F0 (p = 0.002; 0.003). However, no differences were determined between F1-3 and F4-6 Groups (p = 0.158). The sensitivity and specificity values were found as 50 and 97.37% (cutoff 1.361 9 10-3 mm2/s, AUC 0.816, p \ 0.001). Discussion/Conclusion: High ISHAK scores are important data which affect the chronic liver disease. However, biopsy is an invasive procedure and causes complications; therefore, this situation leads to the development of alternative methods (e.g. elastography and fibroscan). In the future, DWI may replace biopsy in determining the scores at the early stages of fibrosis. References: 1. Taouli B, Tolia AJ, Losada M, Babb JS, Chan ES, Bannan MA, Tobias H. Diffusion-weighted MRI for quantification of liver fibrosis:

S509 preliminary experience. AJR Am J Roentgenol. 2007 Oct;189(4):799–806. 2. Bakan AA, Inci E, Bakan S, Gokturk S, Cimilli T. Utility of diffusion-weighted imaging in the evaluation of liver fibrosis. Eur Radiol. 2012 Mar;22(3):682–7. 3. Boulanger Y, Amara M, Lepanto L, Beaudoin G, Nguyen BN, Allaire G, Poliquin M, Nicolet V. Diffusion-weighted MR imaging of the liver of hepatitis C patients. NMR Biomed. 2003 May;16(3):132–6. 4. Soylu A, Kilic¸kesmez O, Poturog˘lu S, Dolapc¸iog˘lu C, Serez K, Sevindir I, Yas¸ ar N, Akyildiz M, Kumbasar B. Utility of diffusionweighted MRI for assessing liver fibrosis in patients with chronic active hepatitis. Diagn Interv Radiol. 2010 Sep;16(3):204–8. 5. Kocakoc E, Bakan AA, Poyrazoglu OK, Dagli AF, Gul Y, Cicekci M, Bahcecioglu IH. Assessment of Liver Fibrosis with DiffusionWeighted Magnetic Resonance Imaging Using Different b-values in Chronic Viral Hepatitis. Med Princ Pract 2015;24(6):522–6.

538 WITHDRAWN 539 Comparison study of synthetic high b value diffusionweighted images using intravoxel incoherent motion model with acquired high b value diffusion-weighted images in diagnosis of prostate cancer S.Y. Kim, I.C. Song, J.Y. Cho, S.H. Kim RADIOLOGY, SEOUL NATIONAL UNIVERSITY HOSPITAL, SEOUL/KOREA, REPUBLIC OF Purpose/Introduction: To evaluate the clinical performance of synthetic high b value diffusion weighted images (DWI) obtained using serial lower b value DWI using intravoxel incoherent motion (IVIM) model and to compare synthetic high b value DWI (s-DWI) to acquired high b value DWI (a-DWI) in diagnosis of prostate cancer. Subjects and Methods: All MR data were acquired using an IVIM technique on a 3T MR unit. A total of five b factors of 0, 50, 150, 300 and 1000 s/mm2 were set in EPI diffusion sequence. Diffusion coefficients were estimated from a single exponential decay diffusion model using all data of five b factors. IVIM parameters were estimated by a two-step method based on the handling of diffusion and perfusion-related region separately. The b value of 300 s/mm2 was used as a threshold b factor in pseudo-diffusion, perfusion-related region. Diffusion in a single exponential decay model, and diffusion, pseudo-diffusion and perfusion fraction parameters in IVIM model were measured using the nonlinear least squares method. Thereafter, synthetic high b value (b = 1500 s/mm2) diffusion-weighted images were obtained by both IVIM and single exponential decay diffusion models. From eight patients’ MR data, synthetic and acquired DWIs with a b value of 1500 s/mm2 were evaluated and compared each other in view of clinical performance and image quality. The contrast ratio, signal to noise ratio (SNR), contrast to noise radio (CNR), lesion detectability and the overall diagnostic acceptability using 5-point scale were evaluated, respectively. The overall diagnostic acceptability was graded on a 5-point scale: 1, non-diagnostic image quality; 2, suboptimal or limited image quality; 3, standard image quality; 4, better than standard image quality and 5, excellent image quality. Results: About same lesions detected as prostate cancer, the contrast ratios between lesions and surrounding parenchyma obtained from s-DWI were superior to those from a-DWI (mean ratios; 0.78, 0.64,

123

S510 respectively). The mean SNR from s-DWI was not inferior to that from a-DWI (4.58, 4.79, respectively). All lesions detected in a-DWI could be detected in s-DWI. The overall diagnostic acceptability of s-DWI showed more than standard diagnostic acceptability (score C3). Discussion/Conclusion: Synthetic high b value DWI obtained using IVIM model showed equal image quality and superior contrast ratio between prostate cancer and surrounding parenchyma to acquired high b value DWI. The performance of synthetic DWI for detection of cancer lesion was comparable to that of acquired DWI. References: Luciani A, Vignaud A, Cavet M, et al. Liver cirrhosis: intravoxel incoherent motion MR imaging–pilot study. Radiology. 2008;249(3):891–9. Grant KB, Agarwal HK, Shih JH, et al. Comparison of calculated and acquired high b value diffusion-weighted imaging in prostate cancer. Abdom Imaging. 2015;40(3):578–86. Mesmann C, Sigovan M, Berner LP, et al. Evaluation of image quality of DWIBS versus DWI sequences in thoracic MRI at 3T. Magn Reson Imaging. 2014;32(10):1237–41. Bittencourt LK, Attenberger UI, Lima D, et al. Feasibility study of computed vs measured high b-value (1400 s/mm2) diffusion-weighted MR images of the prostate. World J Radiol. 2014;6(6):374–80. Maas MC, Fu¨tterer JJ, Scheenen TW. Quantitative evaluation of computed high B value diffusion-weighted magnetic resonance imaging of the prostate. Invest Radiol. 2013;48(11):779–86.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 = 30.2 ± 8.7 ms (Fig. 1). Sodium T1 measurements performed on the vials containing the saline solutions, distributed over different positions within the coil volume, showed a minimal variability of the T1 times (T1 = 22.5 ± 1.8 ms, range = 20.5–23.6 ms).

discs

Discussion/Conclusion: T1 measurement of 23Na signal can be performed in clinical settings using a segmented saturation TrueFISP sequence. Measurements are not significantly affected by B1 and B0 inhomogeneities and fast. This method represents a robust alternative to IR and VFA methods for quantitative sodium MRI. References: 1 Am J Cardiol 2012; 109:1510–1513. 2 MRM 2001; 45:720–723.

540 Saturation Recovery TrueFISP for T1 mapping of sodium signal in the abdomen

541 MRI analysis of endobiliary photodynamic therapy (EPDT) effects in Klatskin tumor patients

A. Ciritsis1, A. Becker1, M. Klarhoefer2, C. Rossi1 1 Department of Diagnostic and Interventional Radiology, University Hospital Zu¨rich, Zu¨rich/SWITZERLAND, 2MRI Research, SiemensHealthcare Switzerland, Zu¨rich/SWITZERLAND

M. Shorikov1, M. Lapteva1, D. Frantsev1, O. Sergeeva1, V. Panov2, B. Dolgushin1 1 Radiological, Research Institute of Clinical and Experimental Radiology, Federal State Institution ‘‘N.N. Blokhin Russian Cancer Research Center’’, Russian Ministry of Health, Moscow/RUSSIAN FEDERATION, 2Radiology, Russian Medical Academy of Continuous Professional Education, Moscow/RUSSIAN FEDERATION

Purpose/Introduction: Quantitative 23Na-MRI potentially allows for non-invasive monitoring of body sodium distribution and of sodiumpotassium adenosine tri-phosphatase activity, which have a high clinical relevance1. Inversion recovery (IR) or variable flip angle (VFA) methods for 23Na-T1 relaxation time quantification are not feasible in all clinical settings, due to limitations of the maximum achievable transmission power and to B1 inhomogeneities. In this study, we propose the use of a saturation recovery (SR) TrueFISP2 sequence for quantification of 23Na-T1 times in the abdomen. Subjects and Methods: Sodium T1 measurements were performed at 3 Tesla using a 23Na surface coil on phantoms and in one healthy volunteer. Sodium signal was acquired using a centric-reordered 3D TrueFISP readout. T1-weighting was achieved by applying a SR nonselective 90 pre-pulse. Sequence parameters were: TR/TE = 6.2 ms/ 3.08 ms; flip angle = 60; bandwidth = 260 Hz/px. The time-delay from the saturation-pulse to the center of k-space acquisition was set to 60–120–200–1000 ms, respectively. Total acquisition time was about 4 min. The longitudinal relaxation (T1) was computed by fitting the MR signal measured at different time-delay to the function Local T1 values were assessed in the volunteer over the renal cortex and medulla, and the intravertebral discs. In each tissue type, mean values and standard deviation were computed over 5 different RoIs. Measurements on phantoms (140 mM NaCl solution) provided an estimate of fit-robustness in dependence of potential B1 and B0 inhomogeneities. Results: In the volunteer, the following T1 values were measured: T1,medulla = 36.3 ± 7.8 ms; T1,cortex = 33.6 ± 3.3 ms; T1,intravertebral

123

Purpose/Introduction: EPDT is a binary tumor destruction modality implying more or less specific accumulation of intravenously injected agent named photosensitizer by proliferative cells and inflammatory tissue following by laser irradiation [1, 2] proposed for survival prolongation in Klatskin tumor patients. MRI-evaluation of EPDT results to our knowledge wasn’t previously published. The purpose of this study is to compare quantitative MRI-parameters in Klatskin tumor patients before and after PDT, and to determine parameters that may be able to predict a negative outcome. Subjects and Methods: Twelve non-surgical Klatskin tumor patients were enrolled in the study. In a period of 2012–2017 they underwent 29 EPDT procedures (interprocedural intervals varies from 3 to 8 months, median-5 months). Patients who died within usual time gap between procedures (two patients that died after 11, 12 months were excluded) comprised an additional ‘non-survivors’ group. 58 MRI-studies performed on 1.5T MRI-scanner within 1 week to 2 months (median 20 days) before the procedure and 0.5–6 months (median-2.7 months) after. 2D T2WI-HASTE fat-saturated, 2D T2WI-BLADE fat-saturated and 2D DWI + ADC were evaluated. Signal intensity (SI) and its standard deviation (SD) in 20–50 px ROI were measured in the tumor tissue. SI of the muscle, spleen and SVIII of the liver were used for normalization purposes. The maximal diameter of the tumor tissue was evaluated on T2WI-BLADEfs.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results: After the first PDT SI (tumor)/SI (muscle) on both types of T2WI (Fig. 1A,B) and ADC (tumor)/ADC (liver) increased (Fig. 1C) in comparison to MRI-study before the procedure (p \ 0.009–0.047), these changes may be connected to edema and inflammation. Tumor size after the first PDT increased (p \ 0.041, fig. 1D). All these changes didn’t have impact on overall survival (p [ 0.1).

S511 High heterogeneity (SD) of the tumor on T2WI-HASTE, ADC, its increase after the procedure on T2WI-BLADE are possible hazardous MR-signs. References: Henderson B.W., Dougherty T.J. How does photodynamic therapy work? PhotochemPhotobiol. 1992; 55 (1): 145–157 doi: 10.1111/j.1751-1097.1992.tb04222.x. PMID: 1603846 [PubMed— indexed for MEDLINE]. Dougherty T.J., Gomer C.J., Henderson B.W., Jori G., Kessel D., Korbelik M., Moan J., Peng Q. Photodynamic therapy. J Natl Cancer Inst. 1998; 90 (12): 889–905. doi: 10.1093/jnci/90.12.889. PMID: 9637138 [PubMed—indexed for MEDLINE].

542 MRI analysis of Gd-BOPTA excretion in patients with compromised and non-compromised bile ducts

After all the PDTs only SI (tumor)/SI (muscle) on T2WI-HASTE increased significantly (p \ 0.014). No changes in tumor diameter were spotted (p [ 0.08). Oddly, after all the procedures increased SI (spleen) was present (p \ 0.004). Comparison of survivors and non-survivors revealed that time since the previous till the next procedure was significantly longer than time since the procedure till death in non-survivors (p \ 0.04, medians-5 and 3 months respectively). SD (after the last procedure)/SD (before the last procedure) on T2WI-BLADE was significantly higher in nonsurvivors (p \ 0.04, Fig. 2A, 0-non-survivors, 1-survivors). Nonsurvivors had more heterogeneous signal (measured as SD) on T2WIHASTE and ADC (p \ 0.03–0.04, Fig. 2, B, C, D, 0-non-survivors, 1-survivors). Logistic regression analysis revealed diagnostic accuracy of 85% of the combination of these potentially hazardous MRIsigns.

M. Shorikov1, P. Polyakov2, D. Frantsev1, O. Sergeeva1, V. Panov3, B. Dolgushin1 1 Radiological, Research Institute of Clinical and Experimental Radiology, Federal State Institution «N.N. Blokhin Russian Cancer Research Center» , Russian Ministry of Health, Moscow/RUSSIAN FEDERATION, 2Medical cybernetics, Russian National Research Medical University named after N.I. Pirogov, Moscow/RUSSIAN FEDERATION, 3Radiology, Russian Medical Academy of Continuous Professional Education, Moscow/RUSSIAN FEDERATION. Purpose/Introduction: Magnetic resonance contrast agent (MRCA) Gd-BOPTA demonstrated dual excretion either by liver or kidneys. As it was previously shown in obstructive jaundice models, a bile duct ligation in the rodents significantly reduced the liver-to-kidneys excretion ratio compared to that in the animals after sham procedure [1]. MRI assessment of the hepatic function based on Gd-BOPTA excretion shift is attempted in the study. Subjects and Methods: Twenty-five patients that had MRI-examination with Gd-BOPTA in period since 2015 till 2016 showing normal total serum bilirubin level were retrospectively included in the study: 15 patients with cholestatic liver diseases (14—Klatskin tumor, 1—primary sclerosing cholangitis) after biliary decompression (group 1) and 10 patients without cholestasis (10—metastatic neuroendocrine tumors, group 2). They were studied on 1.5T MRI-system. MRCP examination was also done in all the cases. T1VIBE-images were acquired before and after 10 ml Gd-BOPTA i.v. injection. Images for arterial, portal, venous and delayed (1, 20, 40 min after injection) phases were acquired; ROIs (about 50 pixels) were selected in liver and kidney. One patient from group 1 had an additional study with 7.5 ml of Gd-DTPA a month later; T2WI, DWI with ADC-maps were also evaluated in this patient for liver fibrosis detection. Results: Endobiliary Gd-BOPTA was registered in 3/10 patients in group 1 and in all patients in group 2. Maximal diameter of the bile duct in a lobe correlated to SI (Liver) measured as percentage of precontrast image signal at 20, 40 min phase in an opposed lobe of the liver (Right: r = 0.6, p \ 0.002; Left: r = 0.5, p \ 0.002). SI (Liver)/SI (Kidney) was significantly lower in group 1 compared to group 2 (p \ 0.02) at 20 (Fig. 1A), 40 (Fig. 1B) min phase.

Discussion/Conclusion: SI increase on T2WI-HASTE after all the procedures, on T2WI-HASTE, T2WI-BLADE, ADC after the first PDT appeared to be normal MR-signs after the procedure;

123

S512

Additionally in the patient that had two studies with different MRCAs: (1) Gd-BOPTA: SI (Liver)/SI (Kidney) for IV-V segment with signs of liver fibrosis was significantly lower than SI (Liver)/SI (Kidney) for VIII segment without fibrosis (p \ 0.05); (2) Gd-DRPA: the ratio for IV and VIII segments were not significantly different (p [ 0.05) and were higher than in case of the Gd-BOPTA injection (p \ 0.05). Discussion/Conclusion: Patients with cholestatic liver diseases demonstrated Gd-BOPTA excretion shift even after biliary decompression and serum bilirubin normalization. References: Kirchin MA, Lorusso V, Pirovano G. Compensatory biliary and urinary excretion of gadobenate ion after administration of gadobenate dimeglumine in cases of impaired hepatic or renal function: a mechanism that may aid in the prevention of nephrogenic systemic fibrosis Br J Radiol 2015.

543 Semi-automatic processing of contrast-enhanced magnetic-resonance imaging of kidneys and adrenals in prognosis of efficiency of renal sympathetic denervation in patients with medicamentous-resistant hypertension N.I. Ryumshina1, A.E. Baev2, P.I. Lukyanenok1, V.F. Mordovin3, W.Y. Ussov1 1 Lab. of Tomography, Institute of Cardiology, Tomsk/RUSSIAN FEDERATION, 2Department of Interventional Radiology, Institute of Cardiology, Tomsk/RUSSIAN FEDERATION, 3Department of Arterial Hypertension, Institute of Cardiology, Tomsk/RUSSIAN FEDERATION

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Purpose/Introduction: Although the transcatheter renal sympathetic denervation has become more and more popular as efficient technique of treatment in medicamentous-resistant arterial hypertension, the MRI and in particular the contrast-enhance MRI are not routinely employed for selection of patients and for prognosis of effect of treatment [1]. We have evaluated the possible role of contrast-enhanced MRI studies of the kidneys in prognosis of clinical effect of the renal denervation treatment of the resistant hypertension. Subjects and Methods: 24 patients (as old as 57.8 ± 8.75 years) with resistant hypertension were included, in everybody the radiofrequency ablation (RFA) of the sympathetic nerves of renal arteries was carried out. The average 24-h monitor of the arterial pressure was as 137–210/57 148 mm Hg. The MRI studies were carried out using MRI scanner ‘‘Titan Vantage’’ (by ‘‘Toshiba Medical Ltd’’, with the field induction 1.5 T) and comprised set of renal axial, coronal and sagittal slices in T2, T1, inflow/outflow and STIR modes with and without fat suppression. After this the intravenous contrast enhancement has been carried out (0.5 M gadoversetamide, 0.2 ml/kg), with repeat acquisition of T1-weighted spin-echo images in 3 min and 15 min after injection. Semiquantitative original software has been employed for isolation and quantification of renal volumes in particular of cortex, medulla and of kidney as total, as well as of suprarenal’s volumes, on both sides. Results: Over the group of patients treated with RFA desympathysation of renal arteries led to significant decrease in arterial pressure for (3–38)/(5–22) mm Hg. Significant negative correlation of postoperation decrease of arterial pressure has been detected with both volume of renal cortex, as {[Decrease in SAP] = 42.7 – 0.35 9 (Vleft kidney)} (r = -0.45, p = 0.0498), and also with mass of the right suprarenal gland as {[Decrease in SAP] = 16.5 – 6.78 9 (Vright suprarenal)} (r = -0.350, p = 0.04934). No decrease was achieved in patients with unilateral renal mass over 200 g and right suprarenal gland over 2 g. Discussion/Conclusion: Therefore the anatomic quantitative data of the pre-operative MRI studies of the kidneys can be employed as prognostic tool in patients referred to the renal denervation treatment of the resistant hypertension, and MRI should be carried out in every patient referred for RFA desympathyzation of renal arteries. References: 1. Kandzari D. E. Catheter-Based Renal Denervation for Resistant Hypertension: Rationale and Design of the SYMPLICITY HTN-3 Trial. Clinical Cardiology. 2012; Vol. 35(9): 528–535.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S513

Abdominal Imaging - Preclinical 544 Oxygen gas challenge to assess placental hypoperfusion in a rabbit model using T2* measurements C. Bertholdt1, M. Dap1, C. Schaaf2, O. Morel1, M. Beaumont2 1 IADI, INSERM, VANDOEUVRE-LES-NANCY/FRANCE, 2CIC-IT, CHRU de Nancy, VANDOEUVRE-LES-NANCY/FRANCE Purpose/Introduction: Intrauterine growth restriction (IUGR) is associated with chronic placental hypoperfusion. Conventional 2D ultrasound do not assess perfusion abnormalities. MRI, imaging modality compatible with pregnancy, can provide functional information about placental perfusion (1). Previous studies have shown that T2* measurements changes during oxygen variation was effective to assess chronic placental hypoperfusion in rat model (2, 3). In this work, we used T2* measurements during oxygen gas challenge on a rabbit model of acute placental hypoperfusion. Subjects and Methods: French animal care committee approved all procedures. Placental hypoperfusion was surgically induced by ligation of uteroplacental vessels on 4 rabbits at day 28 of gestation (28/ 31) in one horn to keep the other one for control. Images were acquired on a 3T MR scanner (Signa, HDxt, GE Healthcare, USA) with a transmit-receive knee coil using a multi-echo gradient-echo sequence (TR = 100 ms, flip angle = 15, FOV = 20 cm, slice thickness = 2 mm, matrix = 256 9 256 and 16 TEs ranging from 2.3 to 47.3 ms). Measurements were performed with 100% of O2 and repeated after 5 min at 20% of O2. Fetus vitality was assessed at the end of the procedure by ultrasound. Images were registered before post-processing. T2* maps were obtained by pixel-wise fitting a monoexponential plus constant curve on signal intensity versus TE. DeltaR2* maps were generated from T2* maps. Mean and standard deviation values were computed from ROIs encompassing the placenta on three consecutive slices. Statistical analysis was performed using Wilcoxon rank test. Results: Ultrasound exploration showed that half of the 38 fetuses were alive at the end of the procedure (Table 1). T2* values were correlated with O2 concentrations (Table 2). Dead fetuses presented DeltaR2* values significantly lower than in the live animal group (Table 3) while the standard deviation remained high compared to those obtained on T2* values. Table 1: Fetuses distribution

Discussion/Conclusion: This is the first time that T2* was measured during gas challenge in the rabbit, whom placentation is close to the human one, and in a context of acute hypoperfusion. Comparison between ligated and control fetuses was corrupted due to the high mortality in the ligated group. Even if deltaR2* seems to be good marker of vitality, time of death is not known and some of the fetuses might have been alive during T2* measurements as indicated by the high variability of DeltaR2*. Further works will focus on protocol improvement to maintain fetus vitality to characterize properly placental perfusion defects with MRI. References: 1. Siauve. AmJObstetGynecol. 2015. 2. Aimot-Macron. EurRadiol. 2013. 3. Chalouhi. Radiology. 2013.

545 MRI investigation of prostate pathology in the PTEN knockout murine model of prostate cancer G. Serrano De Almeida1, N. Baxan2, C. Bevan1 1 Department of Surgery & Cancer, Division of Cancer, Imperial College London, London/UNITED KINGDOM, 2Biomedical Imaging Centre, Imperial College London, London/UNITED KINGDOM Purpose/Introduction: Prostate cancer (PCa) is driven by the androgen receptor (AR) signalling axis and begins with prostatic intraepithelial neoplasia (PIN), progressing to invasive adenocarcinoma and eventually metastatic disease. It is treated with androgen deprivation therapies, to which, in late-stage disease, tumours often become resistant and proliferation occurs in a low androgen environment. Mutation of the PTEN tumour suppressor gene is found in approximately 30% of primary human prostate adenocarcinomas and is commonly implicated in metastatic and treatment-refractory disease. In this study, age effects in transgenic murine models with mono PTEN deletion were investigated through MRI. Subjects and Methods: MRI measurements on PTEN. PB-Cre mice were performed on a 9.4T Bruker BioSpec scanner (AVIIIHD) using a 40 mm quadrature mouse body volume coil (Bruker BioSpin GmbH, Ettlingen, Germany). T2-weighted images in axial, sagittal and coronal planes were acquired using a turbo RARE sequence (TR = 5000 ms, TE = 24 ms in plane resolution 150 9 150 lm2 and 800 lm slice thickness. Apparent Diffusio Coefficient (ADC) data was aquired using a trace Spin Echo sequence (TR = 2000 ms, TE = 17.5 ms, 3 slices). The MR images were respiration triggered to minimize motion artefacts. Results: Volumetric measurements show an increase in volume of the prostate of Pten+/-. Cre+ mice with age, at a faster rate than compared to controls. ADC measurements do not show any significant difference between controls and Pten+/-. Cre+ mice at the same age, but a trend of decrease in the ADC with age can be seen in the Pten+/-.

123

S514 Cre+ mice. There was a significant inverse correlation between volume and diffusion over time in the Pten+/-. Cre+ mice (R2 = 0.98).

Discussion/Conclusion: These preliminary data show that the Pten+/-. Cre+ mouse model has faster growth of prostate with age with an apparent decrease of ADC strongly associated. The hipothesis is that an increase in cellularity in the prostate, potentially due to PIN leads to a decrease in the ADC values. Histological studies are still on going to support these findings, however, as previously shown in the literature, ADC can be a useful marker of disease progression, namely in prostate cancer.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 References: Phin S, Moore MW and Cotter PD (2013) Genomic rearrangements of PTEN in prostate cancer. Front. Oncol. 3:240. doi: 10.3389/fonc.2013.00240. Wang, Shunyou et al. (2003) Prostate-specific deletion of the murine Pten tumor suppressor gene leads to metastatic prostate cancer. Cancer Cell, Volume 4, Issue 3, 209–221. Hill, D. K., Kim, E., Teruel, J. R., Jamin, Y., Widerøe, M., Søgaard, C. D., Størkersen, Ø., Rodrigues, D. N., Heindl, A., Yuan, Y., Bathen, T. F. and Moestue, S. A. (2016), Diffusion-weighted MRI for early detection and characterization of prostate cancer in the transgenic adenocarcinoma of the mouse prostate model. J. Magn. Reson. Imaging, 43: 1207–1217. doi:10.1002/jmri.25087. Esen M, Onur MR, Akpolat N, Orhan I, Kocakoc E. Utility of ADC measurement on diffusion-weighted MRI in differentiation of prostate cancer, normal prostate and prostatitis. Quantitative Imaging in Medicine and Surgery. 2013;3(4):210–216. doi: 10.3978/j.issn.2223-4292.2013.08.06.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Cardiovascular, Breast and Chest Imaging 546 Breast MRI evaluation of focal asymmetric opacities seen on only one mammographic view G. Go¨kalp1, F. Pektas2 1 Radiology, Uludag University Medical faculty, Bursa/TURKEY, 2 Radiology, Uludag University Medical Faculty, Bursa/TURKEY Purpose/Introduction: Breast cancer is the most common invasive cancer in women. Mammography is the primary imaging modality for detection of breast cancer and evaluation of breast lesions but mammography sometimes yield equivocal results despite a thorough diagnostic evaluation. Ultrasonography (US) and magnetic resonance imaging (MRI) are adjunctive imaging modalities when mammographic findings are inconclusive. The aim of this study was to evaluate the findings and the usefulness of breast MRI in cases of focal asymmetric opacities seen on only one mammographic view. Subjects and Methods: Sixty asymmetric opacities of 55 female patients in whom MRI was performed for problem solving for focal asymmetric opacities seen on one mammographic view between September 2014 and November 2015 were included in this retrospective study. MRI images of patients were evaluated according to American College of Radiology (ACR) Breast Imaging Reporting and Data Systems (BIRADS) MRI criteria. Results: 61.6% of cases were evaluated as asymmetric fibroglanduler tissue (BIRADS 1), 25% of cases were evaluated as benign (BIRADS 2), 11.6% of cases were evaluated as suspicious for malignancy (BIRADS 4) and 1.6% of cases were evaluated as highly suggestive of malignancy (BIRADS 5). In BIRADS 4 and 5 cases, 4 (6.6%) malignant lesions were histopathologically verified. MRI had a sensitivity, specificity, positive predictive value and negative predictive value of 100, 94.5, 57.1 and 100% respectively. Discussion/Conclusion: As a result focal asymmetric opacities seen on only one mammographic view were most frequently originate from normal fibroglandular tissue, but benign and malignant pathologies can also cause focal asymmetric opacities seen on only one mammographic view. Breast v MRI is an adjunctive modality when conventional imaging methods were inadequate due to dens breast parenchyma, surgical scars, anatomic variations, lesions located closer to chest wall, sternum and axillary tail and for the diagnosis of multifocal, multicentric breast cancer. References: D’Orsi CJ, Sickles EA, Mendelson EB, et al. American College of Radiology Breast Imaging Reporting and Data Systems Atlas. Virginia: American College of Radiology; 2013. Samardar P, Shaw de Paredes E, Grimes MM, Wilson JD. Focal asymmetric densities seen at mammography: US and pathologic correlation. RadioGraphics 2002; 22: 19–33. Ikeda DM, Andersson I, Wattsgard C, et al. Interval carcinomas in the Malmo Mammographic Screening Trial: radiologic appearance and prognostic considerations. AJR Am J Roentgenol 1992; 159: 287–94. 110. Mendelson EB, Harris KM, Doshi N, Tobon H. Infiltrating lobular carcinoma: mammographic patterns with pathologic correlation. AJR Am J Roentgenol 1989; 153:265–71. 111. Patel MR, Whitman GJ. Negative mammograms in symptomatic patients with breast cancer. Acad Radiol 1998; 5: 26–33. 112. Elmore JG, Armstrong K, Lehmann CD, et al. Screening for breast cancer. JAMA 2005; 293: 1245–6. 113. Boyd NF, Guo H, Martin LJ, et al. Mammographic density and the risk and detection of breast cancer. N Engl J Med 2007; 356: 227–336. 114. Stomper PC, D’Souza DJ, DiNotto PA, Arrendondo MA. Analysis of parenchymal density on mammograms in 1353 women 25–79 years old. AJR Am J Roentgenol 1996; 167: 1261–5. 115.

S515 Stavros AT, Thickman D, Rapp CL, et al. Solid breast nodules: use of sonography to distinguish between benign and malignant lesions. Radiology 1995; 196: 123–34. 116. Rhabar G, Sie AC, Hansen GC, et al. Benign versus malignant solid breast masses: US differentiation. Radiology 1999; 213: 889–94. 117. Kolb TM, Lichy J, Newhouse JH. Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patients evaluations. Radiology 2002; 225: 165–75. 118.

547 Preoperative breast MR imaging kinetic features Using computer-aided diagnosis: association with survival outcome in invasive breast cancer patients original research S.Y. Nam1, E.S. Ko2. 1 Radiology, Gil Medical center, Gachon University of Medicine and Science, Incheon/KOREA, REPUBLIC OF, 2Radiology, Samsung Medical Center, Seoul/KOREA, REPUBLIC OF Purpose/Introduction: To evaluate whether the preoperative breast dynamic contrast-enhanced (DCE) magnetic resonance (MR) imaging kinetic features assessed using computer-aided diagnosis (CAD) can predict the survival outcome in invasive breast cancer patients. Subjects and Methods: The Institutional Review Board approved this retrospective study, and waived the need for informed consent. Between March 2011 and December 2011, 301 women who underwent preoperative DCE MR imaging for invasive breast cancer, with CAD data, were identified. The Cox proportional hazards model was used to determine the association between the kinetic features assessed by CAD and the recurrence-free survival (RFS). The peak signal intensity and kinetic enhancement profiles were compared with the clinical-pathological variables using the Student t test and analysis of variance (ANOVA). Results: There were 32 recurrences during a mean follow-up time of 55.2 months (range 5–72 months). On multivariate analysis, a higher peak enhancement (RFS hazard ratio, 1.004 [95% confidence interval (CI): 1.001, 1.006]; P = .013) on DCE MR imaging and a triplenegative subtype (RFS hazard ratio, 21.060 [95% CI: 2.675, 165.780]; P = 0.004) were associated with a poorer RFS. Higher peak enhancement was significantly associated with a higher T stage, clinical stage, and histologic grade. A higher washout component was associated with a higher histologic grade, triple-negative subtype, and pathologic diagnosis of invasive ductal carcinoma. Discussion/Conclusion: Patients with breast cancer that showed a CAD-derived higher peak enhancement on breast MR imaging had worse RFS. Peak enhancement and volumetric analysis of the kinetic patterns was useful for predicting the tumor aggressiveness. References: Baltzer PA, Freiberg C, Beger S, et al. Clinical MR-mammography: are computer-assisted methods superior to visual or manual measurements for curve type analysis? A systematic approach. Academic radiology. 2009;16(9):1070–6. Lehman CD, Peacock S, DeMartini WB, Chen X. A new automated software system to evaluate breast MR examinations: improved specificity without decreased sensitivity. AJR American journal of roentgenology. 2006;187(1):51–6. Meinel LA, Stolpen AH, Berbaum KS, Fajardo LL, Reinhardt JM. Breast MRI lesion classification: improved performance of human readers with a backpropagation neural network computer-aided

123

S516 diagnosis (CAD) system. Journal of magnetic resonance imaging : JMRI. 2007;25(1):89–95. Dromain C, Boyer B, Ferre R, Canale S, Delaloge S, Balleyguier C. Computed-aided diagnosis (CAD) in the detection of breast cancer. European journal of radiology. Journal of magnetic resonance imaging : JMRI. 2007;25(1):89–95. Song SE, Seo BK, Cho KR, et al. Computer-aided detection (CAD) system for breast MRI in assessment of local tumor extent, nodal status, and multifocality of invasive breast cancers: preliminary study. Cancer imaging : the official publication of the International Cancer Imaging Society. 2015;15:1. Bhooshan N, Giger ML, Jansen SA, Li H, Lan L, Newstead GM. Cancerous breast lesions on dynamic contrast-enhanced MR images: computerized characterization for image-based prognostic markers. Radiology. 2010;254(3):680–90. Meeuwis C, van de Ven SM, Stapper G, et al. Computer-aided detection (CAD) for breast MRI: evaluation of efficacy at 3.0 T. Eur Radiol. 2010;20(3):522–8. Leong LC, Gombos EC, Jagadeesan J, Fook-Chong SM. MRI kinetics with volumetric analysis in correlation with hormonal receptor subtypes and histologic grade of invasive breast cancers. AJR American journal of roentgenology. 2015;204(3):W348–56.

548 Can we predict triple negative breast cancer with MR imaging findings? R. Yılmaz1, Z. Bayramoglu1, R.G. Comert1, Y. Toktas1, H. Karanlik2, N. Cabioglu2. 1 Radiology, Istanbul University, Istanbul Faculty of Medicine, Istanbul/TURKEY, 2general surgery, Istanbul University, Istanbul Faculty of Medicine, Istanbul/TURKEY Purpose/Introduction: Triple-negative (TN) breast cancer accounts for 10%–20% of all breast cancers, which do not express estrogen receptors, progesterone receptors, or human epidermal growth factor receptor 2 (1, 2, 3). It compose a relatively small proportion of all breast cancers, however they are responsible for a comparatively large proportion of breast cancer deaths. MR imaging can provide important information not only on the morphology of the lesions; at the same time, the pathology reflected by the signal intensity characteristics and on the functional evaluation of contrast enhancement. We reviewed imaging findings for 64 patients with TN cancers on MR imaging to determine the imaging characteristics of TN cancers that may allows to predict the diagnosis. Subjects and Methods: Between 2012 and 2016, 384 patients who had been diagnosed as having TN cancer, were reviewed, and 64 patients (mean age: 47.5 years, age range: 28–72 years) who underwent MR imaging were included in this study. MR imaging studies and interpreted the images based on the Breast Imaging Reporting and Data System Atlas 5th edition. Results: The TN cancer was mostly located posteriorly; in 29 patients (45%) lesion was located 1/3posterior and in 10 of them (16%) lesion was located both 1/3posterior and 1/3middle together. Superior-outer quadrant was the most location in 40 patients (63%). Fifty-five (86%) of the 64 cancers showed mass and the other 9 (14%) showed nonmasslike enhancement. On MR imaging, 84.4% of the patients had mass enhancement. The most enhancement patterns were rim enhancement (36%) and heterogenous enhancement (42%) in masses and heterogeneous (56%) in nonmasslike lesions. Mass lesions most commonly had round or oval shape (63.6%) with most frequently irregular margins (51%). MR imaging showed unifocal lesion in 40 (63%) of patients. The early enhacement patterns were medium in 25

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 (% 39) and fast in 22 (%34) with washout delay enhacement pattern in 24 of 55 cases. Intratumoral very high signal in 22 (34%) and high signal intensity in 20 (31%) on T2-weighted images was identified. Discussion/Conclusion: Our results show that triple-negative breast cancers more frequently mass lesion, round-oval shape, irregular mass margin and rimheterogenous enhancement with tendency toward a posterior location. Also, MR findings of TN cancers were high signal intensity on T2weighted images and a unifocal lesion. MR imaging definition of TN cancer can help predict TN cancers at the time of diagnosis as well as early planning of pre-treatment and may help to understand TN cancers biological behavior. References: 1- Foulkes WD, Smith IE, Reis-Filho JS. Triple- negative breast cancer. N Engl J Med 2010; 363(20):1938–1948. 2- Newman LA, Reis-Filho JS, Morrow M, Carey LA, King TA. The 2014 Society of Surgical Oncology Susan G. Komen for the Cure Sym- posium: triple-negative breast cancer. Ann Surg Oncol 2015;22(3):874–882. 3- Kim MJ, Ro JY, Ahn SH, Kim HH, Kim SB, Gong G. Clinicopathologic significance of the basal-like subtype of breast cancer: a compari- son with hormone receptor and Her2/neu-over- expressing phenotypes. Hum Pathol 2006; 37:1217–1226.

549 Ferumoxytol-enhanced magnetic resonance angiography (FeMRA) for the assessment of patients with complex anatomy due for vascular access creation S. Stoumpos1, M. Hennessy2, A. Vesey3, R. Kasthuri2, A. Radjenovic1, P. Mark1, D. Kingsmore3, G. Roditi2 1 Cardiovascular Research Centre, University of Glasgow, GLASGOW/UNITED KINGDOM, 2Radiology, Queen Elizabeth University Hospital, Glasgow/UNITED KINGDOM, 3Renal & Transplant Unit, Queen Elizabeth University Hospital, Glasgow/ UNITED KINGDOM Purpose/Introduction: Conventional vascular imaging techniques are often problematic in kidney disease patients due to associated risks, invasiveness, and imprecision. This is particularly true for patients with complex anatomy or stenoses due to previous central vein catheter (CVC) insertions1 or failed vascular access creations. Ferumoxytol is a superparamagnetic iron oxide preparation that has potential as a magnetic resonance imaging contrast agent in assessing the vasculature. Subjects and Methods: Patients requiring vascular mapping of their thorax and upper arms as part of their assessment before upper limb vascular access creation underwent ferumoxytol-enhanced magnetic resonance angiography (FeMRA) between December 1, 2015 and August 1, 2016. All scans were performed for clinical indications where standard imaging techniques were deemed potentially harmful or inconclusive. Image quality was evaluated in arterial and venous compartments. Results: First-pass and steady-state FeMRA using 4 mg/kg body weight of ferumoxytol (diluted fourfold) as intravenous contrast agent were performed in 18 patients [mean age 61.2 (SD 11.5) years] with previous failed vascular access procedures. Ten patients were predialysis [mean eGFR 12.0 (SD 3.4) mL/min] and 8 were receiving dialysis via a CVC. Good arterial and venous enhancements were achieved in central vasculature, and FeMRA was equally reliable for evaluation of the peripheral vessels. The images allowed precise assessment of the arterial and venous walls, luminal diameter and the presence of stenosis, occlusion, or thrombus formation. Complex

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 central vein occlusions were identified in 6 patients (Figures 1 and 2). All patients completed their studies without adverse events.

S517

550 The contribution of magnetic resonance imaging in the diagnosis of clinically unclear traumatic fat necrosis R. Yılmaz, R.G. Comert Radiology, Istanbul University, Istanbul Faculty of Medicine, Istanbul/TURKEY

Discussion/Conclusion: Our preliminary experience supports the feasibility and utility of FeMRA for vascular mapping in patients with complex anatomy due for vascular access creation, especially those with previous CVC insertions who are at a higher risk for central vein stenosis. References: 1. Schillinger F, Schillinger D, Montagnac R, Milcent T. Post catheterisation vein stenosis in haemodialysis: comparative angiographic study of 50 subclavian and 50 internal jugular accesses. Nephrol Dial Transplant. 1991;6(10):722.

Purpose/Introduction: Fat necrosis is a relatively common benign entity in the breast, resulting from a vascular insult to fat cells. The incidence of the disease is estimated to be 0.6% in the breast, representing 2.75% of all benign lesions (1–4). The reason is unknown in many cases (5). Fat necrosis has long been known to be a great mimic of breast cancer on mammography and sonography (6). Our aim is to evaluate the contribution of magnetic resonance (MR) imaging to clinically uncertain travmatic fat necrosis cases and to describe MR imaging features of fat necrosis of the breast. This is the first study in the literature since MR imaging of only traumatic fat necrosis series was performed. Subjects and Methods: Present study included 16 cases where diagnosis could not be made with certainty on ultrasonography and mamography. Fat necrosis detected with MR imaging was histopathologically proven using US-guided biopsies. MR imaging findings of patients with fat necrosis were classified according to the American College of Radiology BI-RADS Atlas 5th edition. Results: Traumatic FN presented as mass in all patients on MR imaging. Lesions were superficially evaluated in five patients (31%). The shapes of the masses were mostly irregular 8/6 (50%) and round 5/16 (31%). In 10 patients (63%), fat signal was observed in the mass. Edema was seen around the mass in 4 patients (25%). The internal enhancement pattern of masses was heterogenous 9/16 (56%), homogenous 5/16 (31%). Complete enhancement of fat necrosis was seen as the same as partial in 8 patients (50%). Architectural distortion were seen in 5 patients (31%) on MR imaging. Discussion/Conclusion: MR imaging has a wide spectrum of findings for fat necrosis and the appearance is the result of the amount of the inflammatory reaction, the liquefied fat, and the fibrosis. FN is usually isointense to fat elsewhere in the breast, a key to diagnosing with MR imaging. As we are more familiar with the variable MR imaging features of fat necrosis, rather than unnecessary biopsies short-term follow-up may be sufficient-appropriate for confidence in the imaging diagnosis. References: 1. Lee BJ, Adair F. Traumatic fat necrosis of the female breast. Ann Surg 1920;72:188–95. 2. Adair FE, Munzer JT. Fat Necrosis of the female breast: report of one hundred test cases. Am J Surg 1947;74:117–28. 3. Hadfield G. Fat necrosis of the breast. Br J Surg 1930;17:673–82. 4. Pullyblank AM, Davies JD, Basten J, Rayter Z. Fat necrosis of the female breast—Hadfield re-visited. Breast 2001;10:388–91. 5. Chala LF, de Barros N, de Camargo Moraes P, Endo E, Kim SJ, Pincerato KM, Carvalho FM et al. Fat necrosis of the breast: mammographic, sonographic, computed tomography, and magnetic resonance imaging findings. Curr Probl Diagn Radiol. 2004;33:106–26. 6. Kinoshita T, Yashiro N, Yoshigi J, Ihara N, Narita M. Fat necrosis of breast: a potential pitfall in breast MRI. Clin Imaging. 2002;26:250–3.

123

S518

551 Automatic Propagation of Left Ventricular Endocardial Boundary Segmentation in Time-Resolved 3D Cardiac MR

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Figure 3 shows the gold standard and propagated contour in yellow and magenta, respectively (phase 2).

G. Belsley1, J. Tourais2, M. Breeuwer2 1 Dept. of Biomedical Engineering, Eindhoven University of Technology, Eindhoven/NETHERLANDS, 2Clinical Science, Philips Healthcare, Best/NETHERLANDS Purpose/Introduction: Time-resolved 3D Cardiac MR is an emerging technique to address several drawbacks of conventional 2D shortand long-axis cine CMR acquisitions. It aims at accelerating data acquisition with less sensitivity to patient motion, and without the need to plan acquisitions along multiple orientations. Accurate quantification of heart function requires accurate delineation of endocardial boundaries, especially of the left ventricle (LV). We have investigated how well an initial manual LV endocardial segmentation at one particular phase in the cardiac cycle can be propagated to all other phases, for our current 3D cine CMR acquisition method. Subjects and Methods: Free-breathing ECG-gated, time-resolved transversal B-FFE (SSFP) 3D CMR data of a healthy volunteer were acquired on a Philips 1.5T clinical scanner, covering the complete LV and cardiac cycle, using a 3D radial (stack-ofstars) k-space trajectory, which was selected for its inherent patient motion insensitivity. The acquisition parameters: TR/TE 3/2 ms, FA 60, FOV 300 9 300 mm, 13 slices, 17 phases, resolution 1.25 9 1.25 9 8 mm, scan time 3 min and 6 s. We adopted the method of active contouring [1] for endocardial contour propagation, in an implementation similar to that of Hautvast et al. [2]. We extensively explored this multi-dimensional parameter space using our Matlab [3] implementation, in order to first choose appropriate ranges for the parameters. To find the optimal parameter values, the method was then trained with 1728 different parameter settings, using gold standard contours as reference, obtained by averaging contours [4] that were manually delineated by 3 expert users. It was subsequently validated against a gold standard for data outside the training set. Results: Figure 1 reports the average and standard deviation of the absolute distance in mm, averaged over all phases.

The boxplot in Figure 2 shows per phase the absolute distance in mm.

123

Discussion/Conclusion: Figure 2 shows that the absolute distances between the gold standard and the automatic segmentation increase as the heart contracts and lower as the heart expands. Figure 3 shows that the gold standard overlaps the automatic segmentation for the majority of the endocardium, which is also confirmed by the metrics in Figure 1. The automatic segmentation performance metrics are comparable to the variability in the gold standard. We have accomplished accurate automatic propagation of the LV endocardial contour despite the fact that 3D cine CMR has lower contrast and resolution than conventional 2D CMR. References: 1. S. Lobregt and M. A. Viergever, ‘‘Discrete dynamic contour model,’’ IEEE Trans. Med. Imaging, vol. 14, no. 1, pp. 12–24, 1995. 2. G. Hautvast et al., ‘‘Automatic contour propagation in cine cardiac magnetic resonance images,’’ IEEE Trans. Med. Imaging, vol. 25, no. 11, pp. 1472–1482, 2006. 3. MATLAB Release 2016a, The MathWorks, Inc., Natick, Massachussets, United States. 4. V. Chalana and Y. Kim, ‘‘A methodology for evaluation of boundary detection algorithms on medical images.,’’ IEEE Trans. Med. Imaging, vol. 16, no. 5, pp. 642–52, 1997.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

552 Influence of motion artifact in myocardial 1H-MR Spectroscopy Y. Tajima1, M. Hirano1, H. Katsuyama1, J. Shoji1, K. Murata1, K. Saito1, K. Tokuuye1, T. Kaji2, Y. Ichiba2, Y. Komori3 1 Tokyo medical university, Tokyo medical university, Tokyo/JAPAN, 2 Department of Application Service, Siemens healthcare K.K., Tokyo/ JAPAN, 3Diagnostic imaging business area DI research & collaboration, Siemens healthcare K.K., Tokyo/JAPAN Purpose/Introduction: The purpose of this study was to investigate the effect of motion artifact due to cardiac phase on 1H-MR Spectra of myocardium. Subjects and Methods: Sixty consecutive patients who were received Cardiac MR studies to exam previously unexplained cardiomyopathy were included in this research. Analysis of cardiac phase was performed with oine software (Argus function; Siemens Healthcare) for assessment of an internal area of left ventricle on a short-axis of segmented TrueFISP Cine images. Three delay times of cardiac phases (early-systolic, P1; end-systolic, P2; mid-diastolic, P3) were decided according to the cine images and velocity curves of wall motion generated from internal ventricle area data (Figure 1).

S519 Results: Eleven of 60 patients were excluded from this study because they were poor at breath-hold or moved during scan and the data could not been evaluated. The ICCs were 0.742 at P1, 0.876 at P2 and 0.666 at P3, respectively. P1 and P3 showed a good correlation, furthermore P2 showed an excellent correlation (Table 1). There was a significant difference between P2 and P3 (P = 0.009). Bland–altman analysis (Figure 2) showed that limits of agreement (LOA) were the widest at P3 (Repeatability Coefficient, CR = 8.23 9 103) and the narrowest at P2 (CR = 5.20 9 103). These indicated better reproducibility at P2. Table1. Summary of reproducibility analysis of water peak signals on MRspectra at three different cardiac phases. acquisition 1 acquisition 2 mean mean± mean± LOALOAICC p diff. CR×103 SD×103 SD×103 LB×103 UB×103 ×103 P1 57.4±24.1 52.6±25.2 0.74 <0.001 5.04 -2.05 12.1 7.09 P2 55.1±25.6 54.1±23.7 0.88 <0.001 0.79 -4.41 6.0 5.20 P3 58.6±23.1 54.5±25.3 0.67 <0.001 4.51 -3.72 12.7 8.23

Solid line is mean difference. Red dashed lines are the limits of agreement (LOA, mean˘ 2 standard deviations) on each phase. LOA was the widest at P1 (Repeatability Coefficient, CR = 16.34 æ 103) and the narrowest at P2 (CR = 6.35 æ 103). These indicated better reproducibility at P2.

P1 was set at early-systolic phase when reducing speed of ventricle is fastest. P2 and P3 were respectively set at the end-systolic and middiastolic phases when the velocities were around ‘‘0’’, indicating the cardiac motion almost stops for a moment. Single-voxel 1H-MR spectroscopic data at interventricular septum was acquired by a point-resolved spectroscopy sequence with breathhold. The parameters as follows: voxel size, 5 9 7 9 15 mm; TR/ TE, 4 R–R intervals/33 ms; average, 1. A total of 512 data points were obtained using a band width of 1960 Hz. ECG-gated cardiac MR Spectroscopy sequences were performed at three different phases (P1, P2 and P3) and repeated 2 times per patient. Spectra without water suppression were quantied using standard line-tting procedure (MRSpectroscopy; Siemens Healthcare) for area under the curve of water signal at 4.7 ppm. The intra-phase reproducibility of 1H-MRS was evaluated by the intraclass correlation coecient (ICC) between water signals of myocardium repeatedly acquired at the same cardiac phase.

Discussion/Conclusion: On the 1H-MRS of the myocardium, the excitation timing aected the motion artifact. References: 1. van der Meer RW, Doornbos J, Kozerke S, Scha¨r M, Bax JJ, Hammer S, et al. Metabolic imaging of myocardial triglyceride content: reproducibility of 1H MR spectroscopy with respiratory navigator gating in volunteers. Radiology. 2007;245(1):251–7. 2. Gillinder L, Goo SY, Cowin G, Strudwick M, van der Geest RJ, Wang WYS, et al. Quantication of Intramyocardial Metabolites by Proton Magnetic Resonance Spectroscopy. Front Cardiovasc Med. 2015;2:24. 3. Rial B, Robson MD, Neubauer S, Schneider JE. Rapid quantication of myocardial lipid content in humans using single breath-hold 1H MRS at 3 Tesla. Magn Reson Med. 2011;66(3):619–24. 4. Felblinger J, Jung B, Slotboom J, Boesch C, Kreis R. Methods and reproducibility of cardiac/respiratory double-triggered (1)H-MR spectroscopy of the human heart. Magn Reson Med. 1999;42(5):903–10. 5. Gabr RE, Sathyanarayana S, Scha¨r M, Weiss RG, Bottomley PA. On restoring motion-induced signal loss in single-voxel magnetic resonance spectra. Magn Reson Med. 2006;56(4):754–60.

123

S520

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

553 DIAGNOSTIC AND PROGNOSTIC VALUE OF CARDIOVASCULAR MAGNETIC RESONANCE IMAGING IN DILATED CARDIOMYOPATHY

554 Impact of pulmonary artery banding on right ventricular function in magnetic resonance imaging: An animal study

S. Mehra, N. Singla Radiodiagnosis, Pgimer Dr Ram Manohar Lohia Hospital, New Delhi/INDIA

H. Gufler1, S. Niefeldt2, S. Wagner3, C. Yerebakan2 1 Radiology, Martin-Luther University Halle-Wittenberg, Halle (Saale)/GERMANY, 2Cardiac Surgery, University Medical Center, Rostock/GERMANY, 3Neuroradiology, Friedrich Schiller University, Jena/GERMANY

Purpose/Introduction: The purpose of our study was to characterize the presence and extent of myocardial fibrosis in ischemic and nonischemic cardiomyopathies and quantify the severity of chamber dilatation snd left ventricular dysfunction in dilated cardiomyopathy. Subjects and Methods: A total of 30 patients of suspected dilated cardiomyopathy based on clinical features and ECG/Echocardiography findings were included in this prospective observational study. MR was performed in all patients using Gradient cine, T1, T2, STIR sequences and 3D dynamic post gadolinium sequences. Results: CMR could evaluate the morphology of the heart and we found a positive correlation between CMR and echocardiography for assessing severity of chamber dilatation. (p \ 0.00001 and r = 0.726). We calculated the end diastolic volume, end systolic volume, ejection fraction and left ventricular mass in all the patients by CMR. Delayed gadolinium enhancement of the myocardium was present in 19 patients out of the total 30 included in our study. Based on this finding these 19 were characterized as the ischemic dilated cardiomyopathy subgroup. 11 patients out of 30 did not demonstrate late gadolinium enhancement of the myocardium and these were categorized as non-ischemic dilated cardiomyopathy. Wall motion abnormalities were demonstrated by CMR in all patients in our study. 40% patients in our study group who had late gadolinium enhancement had a worse clinical outcome with more incidence of adverse cardiac events in the form of arrhythmias, frequent hospitalizations due to cardiac failure and even death. A significant correlation was established between presence of LGE on CMR and adverse cardiovascular outcome. Discussion/Conclusion: CMR plays an important role in the diagnosis of patients of dilated cardiomyopath, allowing quantification of ventricular dysfunction and detection of myocardial fibrosis. Late gadolinium enhancement of the myocardium identifies myocardial fibrosis. The pattern of late gadolinium enhancement differentiates ischemic from non ischemic dilated cardiomyopathy. The extent of LGE correlates with prognosis. References: Karamitsos TD,Francis JM, Myerson S, Selvanayagam JB, Neubauer S. The role of cardiovascular magnetic resonance imaging in heart failure. J Am Coll Cardiol.2009 Oct6;54(15):1407–24. Bluemke DA. MRI of Nonischemic Cardiomyopathy. AJR AM J Roentgenol 2010 OCT;195(4):935–40.

123

Purpose/Introduction: PURPOSE This experimental study describes the adaptive processes of the right ventricular (RV) myocardium after pulmonary artery banding (PAB) in sheep evaluated by cine cardiac magnetic resonance (CMR), phase-contrast CMR (PCCMR) and conductance catheter. Background: RV failure due to pressure overload is a major cause of death in congenital heart diseases. In order to detect the optimal time for correcting these diseases, noninvasive methods such as CMR are desirable to quantify RV volumes and pulmonary artery (PA) flow, but the role of CMR to monitor adaptive processes after PAB has yet to be established. Subjects and Methods: Seven sheep were subjected to PAB; four nonoperated healthy animals served as controls. Cine CMR and PCCMR were performed 3 months after PAB. RV function was assessed before and 3 months after PAB using a conductance catheter. Results: CMR revealed significantly higher RV masses in the operated group (p \ 0.01). RV ejection fraction was lower in the operated group (p \ 0.01) mainly due to elevated RV end-systolic volume (RV-ESV) (p \ 0.01). The time-to-peak PA flow was longer in the banding group (p \ 0.01) than in the healthy control group. No significant alterations were observed for tricuspid annulus movement parameters. Catheterization-depicted RV output decreased from (p \ 0.01) and the preload recruitable stroke work increased significantly (p \ 0.04) after PAB. Discussion/Conclusion: PAB over 3 months caused RV hypertrophy, increased RV-ESV, induced a left-to-right shift of peak PA flow, and increased myocardial contractility. References: Leeuwenburgh BPJ, Helbing WA, Steendijk P, Schoof PH, Baan J. Biventricular systolic function in young lambs subject to chronic systemic right ventricular pressure overload. Am J Physiol Heart Circ Physiol 2001;281:H2697–704. de Vroomen M, Cardozo RH, Steendijk P, van Bel F, Baan J. Improved contractile performance of right ventricle in response to increased RV afterload in newborn lamb. Am J Physiol Heart Circ Physiol 2000;278:H100–05. Saba SG, Chung S, Bhagavatula S, Donnino R, Srichai MB, Saric M, Katz SD, Axel L. Novel and practical cardiovascular magnetic resonance method to quantify mitral annular excursion and recoil applied to hypertrophic cardiomyopathy. J Cardiovasc Magn Reson 2014;16:35.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S521

555 WITHDRAWN 556 Towards simulation of 3D Phase Contrast imaging of kidney vasculatures A. Klepaczko1, P. Szczypin´ski1, M. Strzelecki1, L. Stefan´czyk2 1 Institute of Electronics, Lodz University of Technology, Lodz/ POLAND, 2Department of Diagnostic Imaging, Medical University of Lodz, Ło´dz´/POLAND Purpose/Introduction: Phase Contrast Angiography (PCA) is one of the current imaging protocols for diagnosis of the blood vessel system [1]. Its attractiveness lies in non-invasiveness and the possibility to quantify blood flow velocity in 3 dimensions. This information can facilitate assessment of a given organ perfusion—an important indicator of a tissue state. This study presents a framework for simulation of the PCA protocol applied to vasculature of the. Our aim is to enable prediction of PCA image appearance for a various imaging parameters. Subjects and Methods: Based on a real CT angiography image, we designed two models of renal arterial trees—a normally appearing structure and one with stenosis in the renal artery. The input image (Fig. 1) was processed by the vessel enhancement algorithm [2] followed by flood fill operation. The extracted raster trees were skeletonized to determine the course of arteries centerlines. Next, vessels radii were estimated along each centerline by finding maximum inscribed spheres along the centerlines. The obtained geometrical description allowed to reconstruct the renal trees in COMSOL—a CFD simulation tool (Fig. 2). The flow boundary conditions were determined based on blood flow velocities measured for the same patient using ultrasound examination. Velocity data, as well as topology of a vessel network feeds our custom MRI simulator [3], previously applied to a digital phantom of the brain vasculature [4]. Results: The results embrace a collection of simulated PCA images. The three direction-sensitive images were combined into one volume by calculating the velocity magnitude in each image voxel (Fig. 3). To evaluate correctness of the framework in modeling the hemodynamics with the MRA simulator, the image intensity values were compared with the blood velocity vector map obtained from the CFD simulations. The mean relative error rate was equal to 0.17 ± 0.04.

123

S522

Discussion/Conclusion: This work enables better understanding of the PCA protocol and allows prediction of its performance for different geometries of realistic renal vessel trees and acquisition parameters. Most importantly, these investigations can be performed without employing expensive diagnostic equipment, extensively used in clinics. Therefore, the designed simulator contributes to wider application of the PCA in clinical practice. Work supported by Polish NCN (UMO-2014/15/B/ST7/05227). References: 1. R.W. Brown, et al. ‘‘Magnetic Resonance Imaging: Physical Principles…’’, Wiley, 2014. 2. A.F. Frangi, et al., MICCAI’98:130–137, 1998. 3. A. Klepaczko, et al., PLoS ONE, 9(4):e93689, 2014. 4. A. Klepaczko, et al., Computer Methods and Programs in Biomedicine, 137:293–309, 2016.

557 Application of non-contrast MR-venography for the evaluation of blood flow in patients with varicose of pelvic veins K. Sevostyanova1, A. Chupakhin2, A. Tulupov2 1 Laboratory of Invasive Medical Technologies, Institute of Chemical Biology and Fundamental Medicine SB RAS, Novosibirsk/RUSSIAN FEDERATION, 2Laboratory of differential equations, Lavrentyev Institute of Hydrodynamics SB RAS, Novosibirsk/RUSSIAN FEDERATION Purpose/Introduction: The varicose of pelvic veins is usually diagnosed on the basis of ultrasonography of pelvis, mostly during gynecological examinations. However, the ultrasound technique in most cases does not allow to reveal the cause of the disease. Application of MRI as an alternative diagnostic method in this respect allows the non-invasive study the structures of the body in vivo thus providing a wide range of scientific and diagnostic approaches for visualization of the vascular system. The aim of the study was to evaluate the applicability of MRI for a qualitative and quantitative assessment of venous blood flow in patients with varicose of pelvic veins.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Subjects and Methods: For the examination the 1.5 T MRI scanner ‘‘Achieva’’ (Philips) was used in this study. 19 patients with primary and secondary varicose of pelvic veins of the II-III degree according to ultrasound examination were involved in this study. The study was performed with a routine protocol MRI for pelvis and abdomen, including of T1- and T2-weighted images and STIR. The non-contrast 3D MR-angiography of the vessels and Quantitative Flow (Q-Flow) phase contrast technique were used for a detailed study of quantitative parameters of venous blood flow. Results: The examination of the patients with varicose veins of pelvis revealed the individual anatomical variability of the venous system. The following groups can be defined according to frequency of occurrence: 1. The most frequently occuring group is the isolated varicose pelvic veins and varicose transformation of the left gonadal vein with the retrograde blood flow through it. 2. The second group consists of more rarely appearing expansion of the right gonadal vein, bilateral expansion of the gonadal veins, extension and doubling of gonadal veins. In patients with secondary varicose of pelvic veins after thrombosis of the iliac veins and inferior vena cava, the gonadal veins in most cases become the main collaterals both in women and men. The characteristics of speed flow in venous blood vessels of the pelvis and abdomen were also measured. Patients with primary varicose of pelvic veins revealed that the flow velocity through gonadal veins in horizontal position tends toward zero, or has a distinct retrograde component. This information is fundamental for determining the prognosis and treatment strategy pelvic viens varicosis. Discussion/Conclusion: Application of above mentioned MRI techniques allows not only qualitatively, but also quantitatively evaluate the characteristics of blood flow without contrast enhancement. The advantages of MRI undoubtedly expand diagnostic possibilities in clinical practice and may supplement and clarify the results of ultrasonography, assesses various venous structures for the patients with varicose of pelvic veins. Such MRI data are important for the choice of tactics for conservative and surgical treatment of this pathology. Thus non-contrast MR-venography can be recommended as a second step of diagnostics after a ultrasonography survey that revealed varicose of pelvic veins. The work was supported by the President of the Russian Federation grant (project #MD-5175.2016.7) for works of MRI, grant of the Russian Foundation for Basic Research (project #16-34-00554 mol_a) regarding clinical part and the Russian Science Foundation (the project #17-11-01156) regarding theoretical part. References: Andrews E.J., Fleischer A.C. Sonography for deep venous thrombosis: current and future applications. Ultrasound Q. 2005, Vol. 21, P. 213–225. Spritzer C.E., Arata M.A., Freed K.S. Isolated pelvic deep venous thrombosis: relative frequency as detected with MR imaging. Radiology. 2001,Vol. 219, P. 521–525. Lindquist C.M., Karlicki F., Lawrence P., Strzelczyk J., Pawlyshyn N., Kirkpatrick I.D. Utility of balanced steady-state free precession MR venography in the diagnosis of lower extremity deep venous thrombosis. AJR Am J Roentgenol. 2010, Vol. 194, N 5, P. 1357–64.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S523

558 Flow sensitization in an inhomogeneous excitation field - MR Fingerprinting approach L. Nun˜ez Gonzalez1, D. Papp2, G. Kotek1, J.A. HernandezTamames1 1 Radiology and Nuclear Medicine Department, Erasmus MC, Rotterdam/NETHERLANDS, 2Department of physics, ELTE, Budapest/HUNGARY Purpose/Introduction: Measuring flow speed with time of flight (TOF) based on signal intensity evolution. Subjects and Methods: Data were acquired by using 2D FSPGR sequence (TR = 10 ms, TE = 1.8 ms, matrix 256 9 256, voxel size = 1.56 9 1.56). The scanned object was a tube with continuous flow at various speeds. The flow was generated by a CompuFlow 1000MR pump (Shelley Medical Images Technologies). The signal intensity through the pipe along the flow direction was analyzed for 5 different flow speeds (8, 16, 24, 32 and 40 cm/s). Bloch-Siegert B1 mapping method was used to determine the excitation flip angle along the pathway of the flow. Time of flight signal evolution was simulated for each speed and compared with measured values. The figures depicts the comparison between measurement and simulated data.

123

S524 Results: The signal intensity evolution along the flow pathway is sensitive to the speed. We found a good correlation between simulated and measured data (see Figures 2, 3). A characteristic deviation between simulation and measurement can be observed away from the inflow point. Discussion/Conclusion: The method allows distinguishing different speeds. The simulation provides clearly separated curves for each speed. The initial evolution is similar to the evolution of the real acquired data. The simulation did not include image artifacts; it assumed plug flow and relied on flip angle measurement on a separate test phantom. Theses experiments are a good proof of concept and promising for flow speed measurements. A more extended simulation of signal evolution is required for a flow sensitized reliable dictionary for MR Fingerprinting.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 References: 1- Kim, S. E., & Parker, D. L. (2012). Time-of-flight angiography. In Magnetic Resonance Angiography: Principles and Applications (pp. 39–50). Springer New York. doi: 10.1007/978-1-4419-1686-0_2. 2 - Paul’s Tofts (2009). Methods for quantitative relaxation parameter mapping: measuring T 1 and T 2. ISMRM Hawaii. 3 - Roberts, C., Little, R., Watson, Y., Zhao, S., Buckley, D. L. and Parker, G. J. M. (2011). The effect of blood inflow and B1-field inhomogeneity on measurement of the arterial input function in axial 3D spoiled gradient echo dynamic contrast-enhanced MRI. Magn. Reson. Med., 65: 108–119. doi:10.1002/mrm.22593. 4 - Brown, R.W., Cheng, Y.C.N., Haacke, E.M., Thompson, M.R. & Venkatesan, R. (2014). Magnetic Resonance Imaging: Physical Principles and Sequence Design. Wiley Blackwell, Hoboken, NJ, USA.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Diffusion Weighted Imaging 559 Evaluation of non-Gaussian diffusion by isotropic diffusion weighting S. Vellmer1, R. Stirnberg2, T. Sto¨cker3, I. Maximov4 1 Theory of complex systems, Bernstein Center for Computational Neuroscience, Berlin/GERMANY, 2DZNE, German Center for Neurodegenerative Diseases, Bonn/GERMANY, 3MR Physics, German Centre for Neurodegenerative Diseases, Bonn/GERMANY, 4 Experimental Physics III, TU Dortmund, Dortmund/GERMANY

S525 b-values and repetitions time. We used 30 directions, TE = 130 ms resulting TA = 10:40 min. Results: Figure 1 presents fitting results of the isotropic diffusion weighting a) MDiso and b) Kiso derived with the FAMED sequence [3, 4], c) and d) as well with third order cumulant expansion and the e), f) MD and mean kurtosis with conventional DKI approach. Figure 2 presents Scatter plots and pearson correlations between MDs and and deviations from Gaussian between metrics derived with fitting of isotropic weighting data and conventional DKI.

Purpose/Introduction: Diffusional kurtosis imaging (DKI) quantifies the deviation from free Gaussian diffusion propagator by expanding the diffusion tensor to higher order [1]. Due to the needs to measure in multiple directions with different diffusion weightings b DKI is a time consuming procedure [2]. Using singel-shot isotropic diffusion weighting (SIDW) the signal attenuation depends on the spatially averaged mean diffusivity (MD) which is the trace of the diffusion tensor [3, 4]. Here we expand the approach of SIDW for fast quantification of non-Gaussian diffusion in vivo at 7T. Subjects and Methods: An underlying assumption of isotropic diffusion weighting is, that the displacement of water molecules can be expressed by a Gaussian probability density. If the water motion is restricted due to cell walls the diffusion differs from the Gaussian. We quantify the deviation of the monoexponential fit function. ln (S/S0) = -bMD + 1/6b2MD2Kiso. where Kiso is the deviation from the Gaussian behavior and MD mean diffusivity. To improve the accuracy we also used the third order cumulant approach from Jensen et al. [5]. ln(S/S0) = -bMD + 1/6b2MD2Kiso + (a - 1)K2isoMD3b3/54. with a = 2/7. We performed measurements on a healthy volunteer at a Siemens 7T scanner who gave written informed consent prior to participation. The study was approved by the local ethical committee. We used the FAMED sequence [3] for SIDW with b-values equal to [0, 500, 1000, 2000] s/mm2, D = 110 ms, TE = 160 ms, TR = 10 ms and 9 acquisitions resulting TA = 7:50 min. Conventional PFGS sequence was used for DKI with the same diffusion time,

123

S526

Discussion/Conclusion: We expanded the conventional fitting model of the diffusion signal for SIDW sequences. We found a moderate correlation to mean kurtosis metrics. Our approach might be a useful for fast isotropic quantification of non-Gaussian diffusion. References: [1] Jensen et al., MRM 53.6 (2005) 1432. [2] Hansen et al. MRM 69.6 (2013) 1754. [3] Vellmer et al., JMR 275 (2017) 137. [4] Vellmer et al., JMR 279 (2017) 1. [5] Jensen et al. Proc. ISMRM 25 (2017).

560 About the dependence of Gaussian diffusion and Kurtosis parameters on SNR in prostate DWIs M.G. Di Trani1, S. Monti2, C. Cavaliere2, M. Aiello2, S. Capuani1 1 Physics, CNR ISC UOS, Sapienza University of Rome, Rome/ITALY, 2 SDN, IRCCS, Naples/ITALY Purpose/Introduction: Diffusion Weighted Imaging (DWI) plays an increasingly important role in diagnosis of prostate cancer, that is the second most common malignancy in men worldwide [1]. The evaluation of signal-to-noise ratio (SNR) is an important step to estimate image quality and the reliability of diffusion-derived parameters, in

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 particular when strong diffusion weightings are employed. In order to avoid DWI signal to fall below the noise floor, Descoteaux et al. [2] suggest SNR [10. However, it is not easy to choose the background region of interest (ROI) on prostate DWI, since there is no visible background air if the FOV is small. Therefore, in literature, the noise is evaluated in different prostate background regions, such as air in rectum[3], obturator internus muscle[4] or air outside subject body. In this work we aim to evaluate the dependence of mean diffusivity (MD), fractional anisotropy (FA), apparent kurtosis (K) and K-derived apparent diffusion coefficient (ADCK) parameters on SNR. Moreover, we show how SNR depends on the b-values, on the number of signal averages (NSA) and also on the choice of the background-ROI. Subjects and Methods: This study was realized on 5 healthy volunteers, following the written informed consent. All MRI examinations were performed with a 3T Siemens Biograph mMR. DWIs were acquired with a Multi-Band Accelerated EPI Pulse Sequences (TE = 86 ms,TR = 3000 ms, vox. size = 1.8 9 1.8 9 3 mm3,multiband acceleration factor = 2), along 6 directions and with 8 different b-values (0, 100, 250, 500, 1000, 1500, 2000, 2500 s/mm2). For each subject, a high-resolution T2-weighted image was acquired with a Turbo Spin-Echo sequence (TE = 112 ms, TR = 4310 ms, voxel size = 0.6 9 0.6 9 3 mm3). ROIs were drawn in central gland (CG), rectum, obturator internus muscle and air on DWIs. SNRs were calculated by dividing the mean signal in CG by the standard deviation (SD) in background; SNR was evaluated for each background-ROI, for each directions and in function of the NSA. MD, FA, K and ADCK maps were calculated.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results: SNR is dependent on the NSA and also on the choice of the background-ROI: the highest SNR is the one evaluated considering the SD of air area. Muscle-ROI provides the worst prostate SNR, instead rectum-ROI manifests an intermediate trend. ADCK and MD value increase with NSA, while K values decrease with NSA, according to Russell Glenn et al. [5].

S527

561 Evaluate the value of DWI sequence in identifying benign and malignant lesion of cervix in comparison with contrast-enhanced method O. Seraydarmansour, G. Jafarinosar MRI, DR. Athari Imaging Center, Teheran/IRAN Purpose/Introduction: Magnetic Resonance Imaging (MRI) is becoming increasingly important in the assessment of gynecologic tumors. Conventional MRI including contrast-enhanced imaging is useful in the differential diagnosis between benign and malignant lesion However, increasing clinical demand for improved lesion characterization has resulted in the introduction of newer sequences such as Diffusion-weighted imaging (DWI). The aim of our study was to investigate the reliability of DWI sequence in the differentiation of malignant from benign lesions of cervix in comparison with contrast – enhanced method. Subjects and Methods: This study included 20 female patients who underwent MR examination with DWI for suspect uterine cervix lesion. MRI was performed using a 1.5 T MR system (Optima 360, GE Healthcare)with a phased array coil. Routine pelvic MR images were acquired as shown in Table 1. Before the MR examination intramuscular administration of Hyoscine was performed to reduce peristalsis (motion artifact). A lubricant gel was injected directly to vagina for all patients. Axial DWI sequences were obtained with b factors 0 and 1000 s/mm2. Dynamic contrast-enhanced fat-suppressed spoiled gradient echo was performed before and immediately after rapid IV injection of 0/12 mmol/kg of contrast material with 15 cc saline at a flow rate of 2 cc/s. This sequence was repeated at 30, 60 and 180 s after injection. Two radiologists with experienced in gynecologic MRI independently evaluated all the sequences performed in each patient. ADC maps were generated by using b0 and b1000 images. For each patient ROI was set by each radiologist within the lesions. We evaluated ADC value for each region.

Discussion/Conclusion: This work demonstrated that the choice of background region and the number of repetitions deeply influence the SNR calculation. Our results show that in order to obtain a SNR [10, 4 averages for low b-values (0, 100, 250, 500 s/mm2) and 6 averages for high b-values (1000, 1500, 2000, 2500 s/mm2) are necessary to extract reliable diffusion parameters. References: [1] Ferlay, J. et al., Int. J. Cancer, (2015): E359–E386. [2] Descoteaux M. et al., IEEE Trans. Med. Imaging (2009); 28: 269–286. [3] Mazaheri Y. et al., Acad Radiol. (2013); 20(8): 1041–1047. [4] Kaji Y. et al., J. Magn. Reson. Imaging (2007);25:517–526. [5] Russell Glenn G. et al., Magnetic Resonance Imaging (2015); 33:124–133.

Results: We assessed the potential malignancy for each lesion by: size larger than 5 cm, cystic or solid-cystic lesion with thick and irregular wall, necrosis, solid components enhancing after administration of contrast media in conventional sequences. Every lesion was analyzed in

123

S528 conventional sequences as well as in DWI sequences with computation of ADC values. ADC values of malignant lesions (1.20 + _22) were significantly lower than benign lesions (1.99_ + 25). Fig 1.

Discussion/Conclusion: In summary, the majority of malignant cervix tumors showed abnormal signal intensity in DWI sequence. On DWI, sensitivity and specificity were 90.5% and 89.1% respectively while on conventional imagingwere 93.2% and 91.6%. References: 1. Pier Paolo Mainenti, Laura Micol Pizzuti, Sabrina Segreto, Marco Comerci, Simona De Fronzo, Federica Romano, VincenzinaCrisci, Michele.2016. Diffusion volume (DV) measurement in endometrial and cervical cancer: A new MRI parameter in the evaluation of thetumor grading and the risk classification. European Journal of Radiology 85, 113–124. [CrossRef].

562 Diffusion MRI to assess the cerebral activation response to fasting status in a glioblastoma mouse model I. Guadilla, M.J. Guille´n, S. Cerda´n Garcı´a-Esteller, P. Lo´pezLarrubia Department of Experimental Models of Human Disease, Instituto de Investigaciones Biome´dicas ‘‘Alberto Sols’’ CSIC-UAM, Madrid/ SPAIN Purpose/Introduction: Glioblastoma (GBM) is a high-grade brain tumor with tendency to infiltrate in the whole brain causing death in a sort time [1]. Our group has previously worked with GBM murine models [2], identifying MRI parameters able to be used as biomarkers of the pathology [3]. On the other hand, we signaled the effects of appetite in human and rodent brains by functional diffusion imaging [4]. In the present work, we wanted to use diffusion MRI to assess a GBM mouse model in fasting conditions. Subjects and Methods: Animal model Adult C57-BL6/J male mice were used in two experimental groups: with glioma cells (GL261) injected in caudate nucleus (n = 7), and without (n = 10). Animals

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 were imaged in two experimental conditions: fed and fasted (16 h). Studies were carried out 21–28 days after injection (when all tumors had a similar volume). Diffusion MRI Anesthetized animals (isoflurane/oxygen) were imaged in a 7T Magnet with a 23 mm volume resonator. Diffusion tensor imaging (DTI) was acquired by applying gradients in 6 directions with the following parameters: D/d = 20/4 ms, TR/TE = 2500/ 43.36 ms, in-plane resolution 0.172 mm/pixel, slice thickness 1.5 mm, axial orientation, and b values of 0, 200, 1000 s/mm2. Images were computed with homemade software to obtain mean diffusivity (MD) and fractional anisotropy (FA). HRMAS Study Mice were euthanized with a high-power microwave and excised brains were regionalized (cortex, prefrontal cortex, hippocampus, thalamus and hypothalamus). 1H HRMAS spectra of biopsies were obtained in a 11.7T Magnet (CPMG sequence, TE = 36 and 144 ms) and analyzed with LCModel. Results: MD values decreased in fasted conditions, in concordance with previous results [4], due to the swelling of astrocytes, but this effect is lower, even statically not significant, in hypothalamus of glioma-bearing mice. On the other side, FA increased in fasted healthy animals, due to higher movement restrictions because of the cells swelling. This increase is again lower and less significant in mice with tumors. Metabolomic alterations were also detected in animals with GBM: decrease in GABA, taurine, N-acetylaspartate, myoinositol and glutamine.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion/Conclusion: We showed that appetite induces changes in MRI diffusion parameters that are associated to the swelling response of astrocyte under fasting conditions. Nevertheless, these changes do not take place, or the effects are lower, in tumor bearing animals, suggesting a down-regulation of the fed/fasted response in this pathology. References: [1] Nader Sanai et al. Neurosurgery 2008, 62:753–66. [2] Borges et al. AJNR 2012, 33(1):24–36. [3] Pe´rez-Carro et al. EJNMMI Res 2014, 4(1):44. [4] Lizarbe et al. Neuroimage 2013, 64:448–57.

563 Does hypertension affect the aging of white matter? Insights from DTI A. Sabisz1, P. Naumczyk2, M. Witkowska2, B. Graff3, K. Jodzio2, D. Gasecki4, A. Konarzewska5, J. Kwela6, E. Szurowska1, K. Narkiewicz3 1 Second Department of Radiology, Medical University of Gdansk, Gdansk/POLAND, 2Institute of Psychology, University of Gdansk, Gdansk/POLAND, 3Department of Hypertension and Diabetology, Medical University of Gdansk, Gdansk/POLAND, 4Department of Hypertension and Diabetology, e Department of Neurology of Adults, Medical University of Gdansk, Gdansk/POLAND, 5Department of Radiology, Medical University of Gdansk, Gdansk/POLAND, 6 Institute of Experimental Physics, University of Gdansk, Gdansk/ POLAND Purpose/Introduction: The aging of the brain manifests itself through cognitive, emotional and social behaviour changes in the elderly. These alternations are underpinned by physiological changes often accompanied by cardiovascular diseases [1], as well as bad nutrition and lifestyle habits. In the hypertensive patients’ population WM microstructure differences derived from DTI studies were reported in the corpus callosum [2, 3]. Additionally the SBP was linearly correlated with decreased regional FA and increased MD, especially in anterior corpus callosum, the inferior fronto-occipital fasciculi and the fibers that project from thalamus to the superior frontal gyrus [3]. The aim of this study was to estimate the changes in the aging brain fibers for each hemisphere separately in patients with hypertension compared to the control group using diffusion parameters obtained by diffusion tensor imaging. Subjects and Methods: One hundred fourty six participants took part in the study : the control group (CON) of 61 healthy individuals and the subjects group (SUB) of 85 hypertensive patients. Age categories were as follows: group I—aged to 47 in CON (N = 22) and 48 in SUB (N = 30); group II—aged from 48 to 56 years in CON (N = 19) and aged from 49 to 58 years in SUB (N = 27); group III— aged over 57 in CON (N = 20) and 59 in SUB (N = 28). MRI examinations were performed on Philips Achieva 3.0T TX scanner with 32-channel head coil. Images of the DTI sequence were used to calculate the values of fractional anisotropy, mean, radial and axial diffusivity in 22 main tracts of the brain. Scheme of the analysis is presented on figure 1.

S529

Results: The two way ANOVA of diffusion coefficients of the nerve pathways, showed an age-related main effect in most of the cases. In contrast, tracts such as: left and right anterior thalamic radiation (in MD, RD, AD), corpus callosum (in MD, RD, AD), left cingulum cingulate gyrus (in MD, AD), left, right uncinate fasciculus (in MD, RD, AD), fornix (in MD, RD, AD), forceps minor (in MD, RD, AD) presented a hypertension-related dependency. Discussion/Conclusion: Age is an important predictor of lowering fractional anisotropy and increasing diffusivity. In patients with hypertension was noted earlier disturbances in microstructure of white matter tracts. This work demonstrated the sensitivity of diffusion parameters to the hypertension in the aging brain. It gives support to the hypothesis that the hypertension stands a risk factor for an acceleration in brain ageing. References: [1] Gasecki D, Kwarciany M, Nyka W, Narkiewicz K, ,,Hypertension, Brain Damage and Cognitive Decline’’ CurrRep. 2013; 15(6): 547–558. [2] Gons RA, van Oudheusden LJ, de Laat KF, van Norden AG, van Uden IW, Norris DG, Zwiers MP, van Dijk E, de Leeuw FE. ‘‘Hypertension is related to the microstructure of the corpus callosum: the RUN DMC study.’’ J Alzheimers Dis. 2012;32(3):623–31. [3] Maillard P, Seshadri S, Beiser A, Himali JJ, Au R, Fletcher E, Carmichael O, Wolf PA, DeCarli C, ‘‘Effects of systolic blood pressure on white-matter integrity in young adults in the Framingham Heart Study: a cross-sectional study’’, Lancet Neurol. 2012 Dec;11(12):1039–47.

564 Monoexpotential and biexpotential fitting in DWI analysis for prediction of the liver fibrosis E. Zawada1, G. Rusak1, M. Moroz2, Z. Serafin1 1 Radiology and Diagnostic Imagine, University Hospital No.1, Bydgoszcz/POLAND, 2Faculty of Health Sciences, Collegium Medicum UMK, Bydgoszcz/POLAND Purpose/Introduction: Diffusion-weighted magnetic resonance imaging (DWI) is a promising tool for the diagnostics of diffuse parenchymal pathologies. Recently, intravoxel incoherent motion (IVIM) model became a subject of studies. In this model biexpotential fitting of data based on an acquisition of multiple b values allows for calculation of perfusion-related diffusion (pseudo-diffusion, D*), perfusion fraction (f), and pure molecular diffusion (D). IVIM allows to separate pure molecular diffusion from microcirculatory part of diffusion signal, being more specific to the physiological processes. Therefore, IVIM has the potential to increase the diagnostic value of DWI in the diagnostics of the liver fibrosis.

123

S530 Purpose: To compare the value of monoexpotential and biexpotential approach to the diffusion-weighted MRI signal in the prediction of the liver fibrosis. Subjects and Methods: Forty patients with hepatitis C were included. Patients were aged 22–75 years with a mean BMI of 24.4. The degree of the liver fibrosis was expressed in Scheuer/Batts-Ludwig/ Tsui scale based on a biopsy. Significant fibrosis was defined as grade [1. Quantification of the apparent diffusion coefficient (ADC) was performed at 1.5 T using 9 b-values (b = 0, 20, 50, 100, 200, 400, 600, 800, 1000 s/mm2). Biexponential fits were applied to diffusion decay curves to calculate pseudo-diffusion coefficient (D*), perfusion fraction (f) and pure molecular diffusion coefficient (D) on the basis of the IVIM model. Monoexponential fits were used to calculate ADCmono (b 0–1000 s/mm2), ADClow (b 0–200 s/mm2), ADCpure (b 400–1000 s/mm2), and ADChigh (b 0, 400–1000 s/mm2). Results: Significant fibrosis was found in 14 subjects. Monoexpotentally-derived ADC parameters were significantly correlated (r

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 0.53–0.99). There was no significant correlation between IVIM parameters. ADCmono and ADChigh were significantly related to the fibrosis grade (P \ 0.02) and none of IVIM parameters presented such an association. ADChigh was the best predictor of significant fibrosis with AUC of 0.81 (95% CI, 0.65–0.92) with sensitivity of 0.57 and specificity of 0.92 for ADC value of B0.0015 (P \ 0.0001). Discussion/Conclusion: In our study group IVIM parameters did not allow for prediction of the liver fibrosis. ADC calculated based on high b values presents considerable specificity in predicting significant fibrosis. References: 1. Galea N, Cantisani V, Taouli B. Liver lesion detection and characterization: Role of diffusionweighted imaging. Journal of magnetic resonance imaging: JMRI. Jun; 2013 37(6):1260–1276. 2. Intravoxel Incoherent Motion MR Imaging for Staging of Hepatic Fibrosis Bin Zhang1,2, Long Liang1,2, Yuhao Dong1, Zhouyang Lian1,2, Wenbo Chen1, Changhong Liang1, Shuixing Zhang1.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Don’t move! Motion Artefcats and Quality Control 565 Geometric accuracy of the MR imaging techniques in the presence of respiratory motion T. Torfeh, R. Hammoud, T. El Kaissi, M. Mcgarry, S. Aouadi, N. Al Hammadi Radiation Oncology, National Center for Cancer Care & Research (NCCCR), Doha/QATAR Purpose/Introduction: MRI is increasingly being used as an alternative modality of 4D-CT for establishing motion management strategies [1–4]. The geometric accuracy of the MR images used for designing these strategies is essential for ensuring the effectiveness of the radiation treatment. In this study, we evaluated the 3D geometric accuracy of balanced steady state gradient echo Cine MR Pulse sequence used in the presence of respiratory motion. Subjects and Methods: A control point based phantom [5] covering a field of view of 300*200*400 mm3 was used. The phantom was placed on top of a track (Fig. 1) which is attached to the synchrony motion table by Accuray (Accuray Inc. Sunnyvale, CA, USA) producing a peak to peak movement of 25 mm. Four ramps of 28 degrees were also used allowing a movement of around 22.1 mm and 11.7 mm in superior/inferior (S/I) and anterior/posterior (A/P) directions respectively.

Results: Geometric distortion was assessed for radial distances of 50, 100 and 150 mm; with a mean geometric distortion of 0.27, 0.41 and 0.55 mm respectively during the S/I motion only. Adding the A/P motion, the mean geometric distortions increased to 0.4, 0.6 and 0.8 mm. Furthermore, blurring was observed in some slices causing a change in the measured FWHM of the control points of around 20%. Discussion/Conclusion: The work presented aimed at designing a procedure allowing the assessment of the geometric accuracy of the MR sequences during motion. Our initial results showed that the mean geometric distortion during motion was within 1 mm, although more predominant in the presence of both S/I and A/P motions. This is a preliminary study; more data is required to assess the influence of imaging parameters on the geometric distortion and its impact on dose calculation. References: 1. Yang J, Cai J, Wang H, et al. Four-Dimensional Magnetic Resonance Imaging Using Axial Body Area as Respiratory Surrogate: Initial Patient Results. International journal of radiation oncology, biology, physics. 2014;88(4):907–912. doi: 10.1016/j.ijrobp.2013.11.245. 2. Hu Y, Caruthers SD, Low DA, Parikh PJ, Mutic S. Respiratory Amplitude Guided 4D Magnetic Resonance Imaging. International journal of radiation oncology, biology, physics. 2013;86(1):198–204. doi:10.1016/j.ijrobp.2012.12.014. 4. Nelson C, Starkschall G, Balter P, Fitzpatrick MJ, Antolak JA, Tolani N & Prado K. Respiration-correlated treatment delivery using

S531 feedback-guided breath hold: a technical study. Med Phys. 2005; 32, 175–181. 5. Tarraf T, Rabih H, Gregory P, Maeve M, Souha A, Azim C, KenPin H, Joseph S, Primoz P, Noora A. Characterization of 3D geometric distortion of magnetic resonance imaging scanners commissioned for radiation therapy planning, Magnetic Resonance Imaging. 2016; Volume 34, Issue 5, 645–653.

566 Wavelet entropy: quantifying small-scale head motion artifacts H. Mattern, A. Sciarra, O. Speck Biomedical Magnetic Resonance, Otto-von-Guericke-Universita¨t Magdeburg, Magdeburg/GERMANY Purpose/Introduction: Prospective motion correction (PMC) can detect and correct small-scale head motion [1, 2] to provide improved image quality. However, quantifying the performance of PMC is not trivial [3]. In this study, wavelet entropy is proposed to quantify artifacts caused by small-scale motion. Subjects and Methods: Wavelet entropy Motion causes noise-like artifacts, blurring, and ghosting. After transforming the image data to wavelets, motion artifacts reduce the sparsity of the wavelet coefficients (similar approach to compressed sensing) compared to an artifact-free image. From the histogram distribution p of the wavelet coefficients the wavelet entropy is computed as (see Fig. 1): EWT = -sum(p log2(p)). Similar to the gradient entropy [3], the wavelet entropy increases with increasing motion artifacts.

Phantom data: In order to generate data with controlled motion artefacts, 3D gradient echo data from a stable phantom were acquired at (0.75 mm)3 resolution while moving the field of view (FoV) randomly with a motion amplitude of 0.00, 0.12, 0.25, 0.37 and 0.50 mm and degrees. The proposed wavelet entropy is compared with the gold standard gradient entropy [3]. In vivo data: Twelve healthy subjects, were scanned (after written consent) with a Nova 32-channel head coil at 7T (Siemens, Erlangen). A 3D MPRAGE sequence with PMC (in-bore camera, motion is corrected each k-space line, more details in [1, 2]) was used to acquire (0.7 mm)3 data within 11:40 min. Retrospective decorrection (DeCo) [4] was used to estimate the corresponding uncorrected volume for each subject. Wavelet entropy and gradient entropy were computed for all PMC and DeCo volumes. Significance was investigated by paired t test. Results: The wavelet entropy increased with the motion amplitude in the phantom experiment while gradient entropy was less sensitive (see Fig. 2). For the in vivo experiment, the observed subject motion and entropy results are shown in Tab. 1. All subjects moved on average less than twice the voxel size (1.4 mm), resulting in minor artifacts (see Fig 3). Decorrection increased wavelet entropy in 8 and gradient entropy in 3 out of 12 cases. While the gradient entropy for PMC-DeCo comparison showed no significant difference (p = 0.23), wavelet entropy revealed a significant difference (p = 0.03).

123

S532

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 HiMR, funded by the FP7 Marie Curie Actions of the European Commission, grant number FP7-PEOPLE-2012-ITN-316716.

567 Improving Robustness of Motion Correction in 3D GRASE PROPELLER Arterial Spin Labeling J. Huber, M. Gu¨nther, M. Vicari MR-Physics, Fraunhofer MEVIS, Bremen/GERMANY Purpose/Introduction: Arterial Spin Labeling (ASL) provides information on tissue perfusion without using a contrast medium [1, 2]. Because ASL is a subtractive technique, it is particularly sensitive to artifacts due to patient motion [3]. This issue can be mitigated by using the selfreferred motion correction algorithm of 3D GRASE PROPELLER (3DGP) [4, 5], where 3D k-space is covered by inplane mutually rotated bricks. The proposed method is aimed to improve the 3DGP technique by reducing its sensitivity to T*2-decay across the in-plane brick width and by providing a more robust motion detection for each brick. Subjects and Methods: In order to prevent fictitious motion estimation due to T*2-decay, the correlation between label and control data is analyzed. Indeed, with no motion, rotation and translation estimates for label and control data must be highly correlated. Moreover, the mean values of the slice-by-slice estimates for rotation and translation are used for motion correction of each brick as a whole. In-vivo measurements of a healthy volunteer were acquired on a 3T Siemens Magnetom Skyra with a twenty-channel head coil and the following parameters: Voxel size = 0.7 9 0.7 9 4 mm3, TR/ TE = 4000 ms/24.5 ms, 10 label/control bricks of 160 9 16 9 8 pixels each, acquired by a 3D GRASE readout with an EPI factor of 16 and a RARE factor of 8. Results: Figure 1 shows that the presence of motion can be associated with label-control correlations of rotation and translation estimates that are below a proper threshold. Figure 2 shows motion estimates that signicantly fluctuate slice-by-slice in each brick, and the corresponding mean values. Figure 3 shows the 3DGP image improvements achieved by using the proposed method. Indeed, label-control correlation thresholds of dR = 0.8 for rotation and dT = 0.9 for translation provide an effective triggering and prevent motion overcorrection due to T*2-decay (cmp image 3e vs. image 3j). Moreover, motion correction based on mean estimates (image 3t) enhances image sharpness in brain outer regions, showing fine structures that well correlate with the reference without motion (image 3j).

Discussion/Conclusion: Wavelet entropy enabled quantification of small-scale motion-related artefacts and was more sensitive than gradient entropy in this study. However, further analysis is required to establish wavelet entropy as a reliable motion metric. References: [1] Maclaren et al. PLoS One 2012. [2] Stucht et al. PLoS One 2015. [3] McGee et al. JMRI 2000. [4] Zahneisen et al. MRM 2015. Acknowledgements: This work was supported by the National Institutes of Health, grant number 1R01-DA021146, and by the Initial Training Network,

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S533 group and the European consortium for ASL in dementia. Magnetic Resonance in Medicine, 73:102–116, 2015. [4] James G. Pipe. Motion CorrectionWith PROPELLER MRI: Application to Head Motion and Free-Breathing Cardiac Imaging. Magnetic Resonance in Medicine, 42(5):963–969, 1999. [5] Huan Tan, W. Scott Hoge, Craig A. Hamilton, Matthias Gu¨nther, and Robert A. Kraft. 3D GRASE PROPELLER: Improved Image Acquisition Technique for Arterial Spin Labeling Perfusion Imaging. Magnetic Resonance in Medicine, 66(1):168–173, 2011.

568 Measuring gradient nonlinearity using field probe array Y.-H. Chu1, Y.-C. Hsu1, F.-H. Lin2, M. Zaitsev1 1 University Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Dept. of Radiology, Medical Physics, Freiburg/GERMANY, 2Institute of Biomedical Engineering, National Taiwan University, Taipei/TAIWAN Purpose/Introduction: Ideal gradient coils produce a magnetic field with its intensity varying linearly with the distance from the magnet iso-center. However, most gradient coils cannot generate a perfectly linear magnetic field within the FOV in practice. Such gradient nonlinearity causes image distortion artifacts, particularly around the FOV margins. Decreasing such distortion is of critical importance for using MRI in interventional procedures [1] and longitudinal quantification of brain morphometry [2, 3]. Different from previous methods [2–5], we propose a new procedure to measure gradient nonlinearity using a field probe array with short imaging time, easy instrument placement and simple algorithm. Subjects and Methods: A 3D field probe array with 24 probes was used in this study. Probes were arranged in a cubic grid with approximately 10 cm distance between probes (Figure 1A). The center of the probe array was placed at 3 9 3 9 3 different positions, each of which was a combination of +3, 0, or -3 cm away from the iso-center in x, y, or z direction. The signal was measured at ±20 mT/ m for x, y, and z gradient with TR = 1 s using 3T scanner (Prisma, Siemens). The scan time is 54 s for all the data used in this study. The magnetic field generated by each gradient coil was modeled by a 5thorder spherical harmonics:

Discussion/Conclusion: The proposed method delivers improved image quality and is promising to enable the use of ASL in organs other than brain, like kidney, where motion is a more severe issue. As a future perspective, three-dimensional correction for 3DGP will be investigated. References: [1] Matthias Gunther. Perfusion Imaging. Journal of Magnetic Resonance Imaging, 40:269–279, 2014. [2] Xavier Golay and Matthias Gunther. Arterial spin labelling: final steps to make it a clinical reality. Magnetic Resonance Materials in Physics, 25:79–82, 2012. [3] D. C. Alsop, J. A. Detre, X. Golay, M. Gunther, J. Hendrikse, L. Hernandez-Garcia, H. Lu, B. J. MacIntosh, L. M. Parkes, M. Smits, M. J. P. van Osch, D. J. J. Wang, E. C. Wong, and G. Zaharchuk. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study

The coefficients of these spherical harmonics were estimated by minimizing the following equation:

To demonstrate the efficacy of our method, we placed another 8ch field probe array randomly at 11 positions and calculated the standard deviation of the estimated probe positions after rigid transformation to the first position. Results: Figure 1B shows the estimated distance that voxel moves because of gradient nonlinearity on three orthogonal views. The distance was close to zero near the iso-center and gradually increased toward the FOV periphery. Figures1CD show the 8 probe locations at 11 positions (88 dots in total). The dots spread wider without the

123

S534 gradient nonlinearity correction. The standard deviation of estimated probe positions was 0.43 mm without correction and 0.09 mm after nonlinear correction.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

569 Susceptibility effect comparison using an MRI depth electrode phantom S. Solis1, R. Martin1, F. Vazquez1, O. Marrufo2, A. Rodriguez3 1 Department of Physics, Faculty of Sciences, UNAM, Mexico City/ MEXICO, 2Department of Neuroimage, INNN MVS, Mexico City/ MEXICO, 3Department of Electrical Engineering, UAM Iztapalapa, Mexico City/MEXICO Purpose/Introduction: Depth electrodes degrade the image quality by metal artifacts in postimplantation magnetic resonance imaging [1]. These unwanted artifacts alter the accurate localization of the electrodes and, the optimal visualization of brain anatomy. Determination of this artefact may offer information on ways to try to reduce it. We measured the susceptibility artifact size using an MRI depth electrode phantom for two common pulse sequences. Subjects and Methods: We used a phantom described in [2] to study the susceptibility artifact size caused by a Spencer probe depth electrode (Medical Instrumentation Co, WI, USA). Fig. 1.a) shows a photo of the agar-fixed electrode phantom. We acquired T2-weighted phantom images using two common pulse sequences for comparison: a) Gradient Echo (FLASH: TR/TE = 9110/115 ms, FOV = 132 9 132 mm2, slice thickness = 2 mm, NEX = 1) and, b) Spin Echo (TSE: TR/TE = 450/26 ms, FOV = 103 9 103 mm2, slice thickness = 3 mm, NEX = 1). Measurements were performed in a 1.5 T Siemens MR imager (Siemens Healthcare Diagnostics, Munich, Germany). These images were used to measure the artifact size together with the software tool SAM (MR-susceptibility Artifact Measurement: sam-toolbox.sf.net) [3].

Discussion/Conclusion: We used a field probe array to estimate the gradient nonlinearity and this information can reduce 79% distortion (0.43 mm–[0.09 mm) in our experiment. There still existed some distortion after the gradient nonlinearity correction, we expect this error can be further reduced using spherical harmonics of a higher order with denser measurements. Compared with previous study in different systems [2, 3], we found less distortion caused by gradient nonlinearity in this study. References: 1. Pappas, E.P., et al., Characterization of system-related geometric distortions in MR images employed in Gamma Knife radiosurgery applications. Physics in Medicine and Biology, 2016. 61(19): p. 6993–7011. 2. Caramanos, Z., et al., Gradient distortions in MRI: Characterizing and correcting for their effects on SIENA-generated measures of brain volume change. Neuroimage, 2010. 49(2): p. 1601–1611. 3. Jovicich, J., et al., Reliability in multi-site structural MRI studies: Effects of gradient non-linearity correction on phantom and human data. Neuroimage, 2006. 30(2): p. 436–443. 4. Weavers, P.T., et al., Image-based gradient non-linearity characterization to determine higher-order spherical harmonic coefficients for improved spatial position accuracy in magnetic resonance imaging. Magnetic Resonance Imaging, 2017. 38: p. 54–62. 5. Doran, S.J., et al., A complete distortion correction for MR images: I. Gradient warp correction. Physics in Medicine and Biology, 2005. 50(7): p. 1343–1361.

123

Results: T2-weighted phantom images were acquired for both pulse sequences. Fig. 1.b, c show the images acquired with FLASH and TSE, respectively. The size of the artifacts was measured along the green line for the two images as shown in Fig. 2.b. Then, profiles as a function of the position were calculated and plotted and shown in Fig. 2.a. Blue (Spin Echo) and green (Gradient Echo) lines represent the artifact size produced by the two sequences.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S535 (2) new method: the first data sample was acquired 20 microseconds prior to the start on the gradient; the trajectory started from the kspace center 20 microseconds before turning on the gradient. Images were reconstructed using the conjugate gradient method [5]. Results: Figures1BCD-1 show the designed x-gradient waveform of one spiral leaf at different time scales. Note that the gradient turned on at time 0. Figures1BCD-2 show the designed and estimated kx coordinates. After probe calibration, the kx was already around 2 rad/ m at time 0. This kx offset cannot be measured in the conventional data readout (kx started from 0).

Discussion/Conclusion: The FLASH image (Fig. 1.b) shows a strong loss of signal produced by the electrodes, whereas the TSE image (Fig. 1. c) allows us to relatively identify the electrodes. In general, the FLASH green profile has samller artifact size along the electrode (green line) compared to the TSE one. There are certain points where the size of the artifact is the same. This can not be appreciated by simple imageinspection, so a specific test shuold be run. These results contradict our initial investigation reported in [2]: the susceptibility artifact is not only signal loss (Fig. 1.b). There is seems to be a trade off between these pulse sequences: FLASH images allow us anatomy visualization with large susceptibility artifacts. TSE images produce small size artifacts but do not permit to easily visualize the anatomy because of the signal loss. References: 1. van Rooijen, B. D, et. al. Neurosurgery 73(3) (2013): 543–549. doi: 10.1227/01.neu.0000431478.79536.68. 2. Garcia, J., et al. AIP Conference Proceedings. 1494 (1) 2012. doi: http://dx.doi.org/10.1063/1.4764610. 3. F. Gu¨ttler, et. al. Biomed Tech 2012; 57 (Suppl. 1) 2012. doi 10.1515/bmt-2012-4161.

570 Correcting pre-emphasis artifacts in interleaved spiral imaging Y.-C. Hsu1, Y.-H. Chu1, M. Zaitsev1, F.-H. Lin2 1 University Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Dept. of Radiology, Medical Physics, Freiburg/GERMANY, 2Institute of Biomedical Engineering, National Taiwan University, Taipei/TAIWAN Purpose/Introduction: Spiral imaging is a fast imaging method that uses a smooth k-space trajectory to efficiently sample 2D k-space [1]. Due to eddy currents and system imperfections, high-quality spiral imaging requires careful calibration of gradient waveforms. Different from estimating gradient waveforms using pulse sequences [2], one can use field probes to accurately measure gradient waveform along with high-order magnetic field fluctuations [3]. Here we report artifacts in a phantom experiment using interleaved spiral acquisitions. This artifact was related to ignoring the gradient moment accumulated before turning on the designed magnetic field gradients. Subjects and Methods: Eight-leaf interleaved spiral imaging sequence (TE/TR = 5 ms/1000 ms, FOV = 200 mm 9 200mm,slice thickness = 2 mm, resolution = 1 mm 9 1mm) was implemented on a 3T MRI system (Prisma, Siemens). A twenty-four channel planar field probe array was used to estimate the 0th and the 1st order magnetic fields [4]. A spherical phantom and a human volunteer were imaged. Two k-space trajectories and signal sampling methods were used for image reconstruction: (1) conventional method: first data sample was acquired exactly at the moment of turning on the gradient; the trajectory started from the k-space center.

Figure 2 shows the reconstructed phantom images. A ring artifact was observed in the uncorrected image (red arrow).

Figure 3 shows the reconstructed leg image with zoomed-in views. Higher noise in the background and bone region (red arrows) was

123

S536 found in uncorrected images. These artifacts were greatly reduced using new sampling method. The residue artifacts were caused by the unsaturated fat signal.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion/Conclusion: Ignoring the gradient moment accumulated before turning on the gradient can reduce the image quality. This artifact was caused by the reconstruction assuming that the k-space trajectory for each spiral leaf to start from the k-space center at the moment of turning on the gradient. Our probe measurements disproved this assumption in reality. Extending the readout time may be important for other center-out k-space trajectory, such as 3D radial UTE sequences. References: 1. Glover, G. Basic and advanced concepts of spiral imaging. in International Society for Magnetic Resonance in Medicine Fast MRI Workshop, Asilomar, CA. 1997. 2. Duyn, J.H., et al., Simple correction method fork-space trajectory deviations in MRI. 1998, Elsevier. 3. Barmet, C., et al., A transmit/receive system for magnetic field monitoring of in vivo MRI. Magnetic resonance in medicine, 2009. 62(1): p. 269–276. 4. Chu, Y., Y. Hsu, and F. Lin, Spiral imaging trajectory mapping using high density 25-channel field probe array. Proceedings of the 23th Annual Meeting of ISMRM, 2015: p. 1014. 5. Wilm, B.J., et al., Higher order reconstruction for MRI in the presence of spatiotemporal field perturbations. Magnetic resonance in medicine, 2011. 65(6): p. 1690–1701.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S537

fMRI 571 Toward WM CBF dynamics characterization during fMRI L. Greco, O. Reynaud CIBM, EPFL, Lausanne/SWITZERLAND Purpose/Introduction: White matter fMRI1-3 remain controversial due to low sensitivity (lesser blood flow compared to gray matter4) and poorly understood fMRI mechanisms. Here, we use Arterial Spin Labeling (ASL), a non-invasive, quantitative, reproducible fMRI technique5, 6, to investigate Cerebral Blood Flow (CBF) dynamics in white/gray matter (WM/GM) during finger-tapping, a task during which WM BOLD fMRI was previously reported2, 3 ,7. Subjects and Methods: For ASL, a FAIR-QUIPSSII8, 9 2D-EPI sequence used recommended parameters10: TR/TE = 3000/17 ms, res. 3 9 3 9 5 mm—6 slices, TI1/TI2 = 700/1800 ms. fMRI runs consisted in 125 {control/tag} pairs (12 min, 33 s). CBF was estimated11,12 using T1,blood = 2170 ms13 and T2 9 (tissue/blood) = 25/ 12 ms14 at 7T. Pilot studies confirmed WM CBF could be measured voxel-wise as in (15, 16) and CBF changes detected at ROI level during finger-tapping7 (Fig. 1).

ASL data was acquired during finger-tapping (12 blocks 9 30 s 9 [ON/OFF]) on 6 subjects (11 runs). A GLM analysis was run on motion-corrected, smoothed perfusion-weighted data using spm1217. For ROI analysis in WM, CBF and BOLD signal time-courses were averaged over all blocks to increase sensitivity. Enhanced ASL was acquired on two additional subjects using optimized background suppression18 (BS = ON/OFF) and ASL inversion pulses19,20 (IR = HS/TR-FOCI). Three runs were acquired with IR/ BS = [HS/OFF||TR-FOCI/OFF||TR-FOCI/ON]. Results: Conventional GLM analysis on perfusion-weighted data resulted in very few activated voxels in the corona radiata, all below the FWE threshold. GM BOLD/CBF activation clusters and timecourses were found robust and reproducible within subjects (Fig. 2).

Without BS, pronounced image artefacts resulted in poor tSNR in WM (Fig. 3A) and unusable WM CBF time-courses. Using BS + TR-FOCI pulses significantly reduced image artefacts, and increased tSNR overall (Fig. 3A) and in WM (tSNR = 0.5 ± 0.1). Most WM perfusion changes observed at ROI level (whole corona radiata) occured at the stimulation onset/offset (Fig. 3B, arrows).

Discussion/Conclusion: Regarding WM CBF, the absence of elevated plateau suggests negligible partial volume effects, and different dynamics vs. GM. Using BS, little CBF variations during rest/stimulation indicated efficient cancellation of physiological signals contribution between blocks. Despite efficient BS (*70–75%) and small TE, remaining BOLD signal contributions could arise. Further developments will include improved BS using 3D imaging and image acceleration, refined ROIs, increased statistical power, jittering of the onset/duration of stimulation, and careful physiological signal monitoring/removal21,22. Poor tSNR prevented the reliable characterization of WM CBF dynamics using conventional ASL. Efficient BS18 and inversion efficiency19,20 are essential to enhance ASL specificity at high field. References: 1. Gawryluk JR, Mazerolle EL, D’Arcy RCN. Does functional MRI detect activation in white matter? A review of emerging evidence,

123

S538 issues, and future directions. Front. Neurosci. 2014. doi: 10.3389/fnins.2014.00239. 2. Gawryluk JR, Mazerolle EL, Brewer KD, Beyea SD, D’Arcy RCN. Investigation of fMRI activation in the internal capsule. BMC Neurosci. 2011;12:56. doi:10.1186/1471-2202-12-56. 3. Mazerolle EL, Beyea SD, Gawryluk JR, Brewer KD, Bowen C V, D’Arcy RC. White matter fMRI activation in the internal capsule: Colocalization with DTI tractography. Neuroimage 2010;50:616–621. 4. Helenius J, Perkio¨ J, Soinne L, Østergaard L, Carano R a D, Salonen O, Savolainen S, Kaste M, Aronen HJ, Tatlisumak T. Cerebral hemodynamics in a healthy population measured by dynamic susceptibility contrast MR imaging. Acta radiol. 2003;44:538–546. doi:10.1080/j.1600-0455.2003.00104.x. 5. Williams DS, Detre JA, Leigh JS, Koretsky AP. Magnetic resonance imaging of perfusion using spin inversion of arterial water. Proc. Natl. Acad. Sci. U. S. A. 1992;89:212–6. doi: 10.1073/pnas.89.9.4220e. 6. Detre J a, Wang J, Wang Z, Rao H. Arterial spin-labeled perfusion MRI in basic and clinical neuroscience. 2009. doi: 10.1097/WCO.0b013e32832d9505. 7. Greco L, Reynaud O. Arterial Spin Labeling fMRI in White Matter at 7 Tesla. In: Proc. Intl. Soc. Magn. Reson. Med.; 2017. p. 5253. 8. Kim SG. Quantification of Relative Cerebral Blood-Flow Change by Flow-Sensitive Alternating Inversion-Recovery (FAIR) Technique—Application to Functional Mapping. Magn. Reson. Med. 1995;34:293–301. doi:10.1002/mrm.1910340303. 9. Wong EC, Buxton RB, Frank LR. Quantitative imaging of perfusion using a single subtraction (QUIPSS and QUIPSS II). Magn. Reson. Med. 1998;39:702–708. doi: 10.1002/mrm.1910390506. 10. Alsop DC, Detre JA, Golay X, et al. Recommended implementation of arterial spin-labeled Perfusion mri for clinical applications: A consensus of the ISMRM Perfusion Study group and the European consortium for ASL in dementia. Magn. Reson. Med. 2015;73:102–116. doi:10.1002/mrm.25197. 11. Buxton RB, Frank LR, Wong EC, Siewert B, Warach S, Edelman RR. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn. Reson. Med. 1998;40:383–396. doi: 10.1002/mrm.1910400308. 12. C¸avus¸ ogˇlu M, Pfeuffer J, Ugˇurbil K, Uludagˇ K. Comparison of pulsed arterial spin labeling encoding schemes and absolute perfusion quantification. Magn. Reson. Imaging 2009;27:1039–1045. doi: 10.1016/j.mri.2009.04.002. 13. Grgac K, Van Zijl PCM, Qin Q. Hematocrit and oxygenation dependence of blood 1H2O T1 at 7 T. Magn. Reson. Med. 2013;70:1153–1159. doi:10.1002/mrm.24547. 14. Yacoub E, Shmuel a, Pfeuffer J, Van De Moortele PF, Adriany G, Andersen P, Vaughan JT, Merkle H, Ugurbil K, Hu X. Imaging brain function in humans at 7 Tesla. Magn. Reson. Med. 2001;45:588–94. doi:10.1002/mrm.1080. 15. Gardener AG, Jezzard P. Investigating white matter perfusion using optimal sampling strategy arterial spin labeling at 7 Tesla. Magn Reson Med 2014;0:0–5. doi:10.1002/mrm.25333. 16. Skurdal MJ, Bjornerud A, Van Osch MJP, Nordhoy W, Lagopoulos J, Groote IR. Voxel-Wise Perfusion Assessment in Cerebral White Matter with PCASL at 3T; Is It Possible and How Long Does It Take? PLoS One 2015;10:e0135596. 17. Friston KJ, Holmes a. P, Worsley KJ, Poline J-P, Frith CD, Frackowiak RSJ. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 1995;2:189–210. doi: 10.1002/hbm.460020402. 18. Ye FQ, Frank JA, Weinberger DR, McLaughlin AC. Noise reduction in 3D perfusion imaging by attenuating the static signal in arterial spin tagging (ASSIST). Magn. Reson. Med. 2000;44:92–100. doi:10.1002/1522-2594(200007)44:1\92::AID-MRM14[3.0.CO; 2-M.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 19. Hurley AC, Al-Radaideh A, Bai L, Aickelin U, Coxon R, Glover P, Gowland PA. Tailored RF pulse for magnetization inversion at ultrahigh field. Magn. Reson. Med. 2010;63:51–58. doi: 10.1002/mrm.22167. 20. Warnking JM, Pike GB. Bandwidth-modulated adiabatic RF pulses for uniform selective saturation and inversion. Magn. Reson. Med. 2004;52:1190–1199. doi:10.1002/mrm.20262. 21. van der Zwaag W, Jorge J, Butticaz D, Gruetter R. Physiological noise in human cerebellar fMRI. MAGMA 2015;28:485–492. doi: 10.1007/s10334-015-0483-6. 22. Jorge J, Figueiredo P, van der Zwaag W, Marques JP. Signal fluctuations in fMRI data acquired with 2D-EPI and 3D-EPI at 7 Tesla. Magn. Reson. Imaging 2013;31:212–220. doi: 10.1016/j.mri.2012.07.001.

572 Differential Hemispheric Brain Network Reorganization in two profiles of High IQ Children: a resting-state fMRI study I. Suprano1, C. Delon-Martin2, G. Kocevar1, C. Stamile1, O. Revol3, F. Nusbaum4, D. Sappey-Marinier1 1 CREATIS, Universite´ Claude Bernard Lyon1, Villeurbanne/ FRANCE, 2Grenoble Institut des Neurosciences (GIN) U 1216, ˆ pital Universite´ Grenoble Alpe, La Tranche/FRANCE, 3Ho Neurologique, HCl, Bron/FRANCE, 4Laboratoire P2S & Centre Psyrene, Universite´ Claude Bernard Lyon 1, Lyon/FRANCE Purpose/Introduction: High Potential (HP) children are usually characterized by better cognitive abilities quantified by a high Full Scale Intelligence Quotient (FSIQ [ 130) as measured by the Wechsler Intelligence Scale for Children (WISC-IV). However, some HP children may present different disabilities in managing their attention, emotions, and relationships. These clinical observations have led us to define two profiles based on WISC scores, namely the Homogeneous-HP profile (Hom-HP) and the Heterogeneous-HP profile (Het-HP), the latest being characterized by a significant difference between verbal comprehension index (VCI) and perceptual reasoning index (PRI). Based on our previous diffusion MRI study1, showing hemispheric differences between Het-HP and Hom-HP children, we explored their brain network functional connectivity and topological organization, using a graph-based analysis of resting-state functional MRI (rs-fMRI). Subjects and Methods: Fifty-seven children (13 healthy controls (HC), 24 Het-HP and 20 Hom-HP), (age: 10.1 ± 1.2 years) were scanned with a 1.5T Siemens MRI system. The rs-fMRI data consisted in 250 BOLD-weighted images recorded during 10.3 min of rest that were preprocessed using SPM12 software (data artifacts rejection, head motion correction and normalization to MNI template). Time-series were extracted using Conn Toolbox from a total of 132 regions from Harvard–Oxford Atlas. Connectivity matrices were computed in the [0.025–0.05 Hz] frequency-band using wavelets2, and graphs were extracted at 20% cost. Nodal metrics reflecting network integration (Degree ‘D’), segregation (Clustering Coefficient ‘CC’, Local Efficiency ‘LE’) and hubs (Betweenness Centrality ‘BC’) were estimated using the Brain Connectivity Toolbox3. Network reorganization was evaluated using the hub disruption index (k)4 for each subject, each hemisphere (kRH, kLH) and metric (kD, kCC, kLE and kBC) (Image1). Potential network reorganization between brain hemispheres of HP groups was tested using ANOVA, regressing for age and gender effects.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Results: Analysis of variance revealed several network reorganizations in Het-HP and Hom-HP children (Table 1). Compared to Controls, the Het-HP group showed a large clustering reorganization in the left hemisphere as shown by significant changes in four metrics (kD, kCC, kLE, kBC) and only one in the right hemisphere (kBC). In contrast the Hom-HP group shows no changes except for kCC in the left hemisphere.

S539 Discussion/Conclusion: This study suggested that clinical behavioral differences observed in Het-HP compared to Hom-HP children is supported by a different brain network topological organization related to segregation and integration that occurs mainly in the left hemisphere. These functional results extend our previous findings obtained with diffusion MRI1 showing that Het-HP are characterized by increased connectivity mainly located in the left hemisphere. References: 1. Nusbaum F, Hannoun S et al., Hemispheric Differences in White Matter Microstructure between two profiles of children with high intelligence, Front. In NeuroSci. 2017. 2. Achard S, Salvador R et al. A Resilient, Low-Frequency, SmallWorld Human Brain Functional Network with Highly Connected Association Cortical Hubs. J Neurosci. 2006;26(1):63–72. 3. https://sites.google.com/site/bctnet/. 4. Achard S, Delon-Martin C et al. Hubs of brain functional networks are radically reorganized in comatose patients. PNAS. 2012;109(50):20608-20613.

123

S540

Image Reconstruction and Processing 573 SENSE Implementation on Graphical Processing Unit (GPU) using LU Decomposition I. Ullah, M. Ammar, H. Omer Electrical Engineering, COMSATS Institute of Information Technology, Islamabad/PAKISTAN

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results: Experiments are performed on phantom, simulated brain and two human head data sets acquired from 1.5T and 3T MRI scanners at St. Mary’s Hospital, London. The images are retrospectively undersampled for acceleration-factor 2. Reconstructions are performed using both CPU (Core i7 3.60 GHz) and GPU (Nvidia GTX 780) implementations of the proposed method. The reconstructed images are shown in Figure-2. A comparison of the computation time of the proposed method on CPU and GPU is presented in Table-1.

Purpose/Introduction: SENSE is a Parallel Magnetic Resonance Imaging (pMRI) technique that utilizes receiver coil sensitivities to reconstruct MR images from the acquired under-sampled k-space data [1]. SENSE algorithm requires the inversion of rectangular encoding matrices as part of image reconstruction. Methods such as Cholesky decomposition [3], QR decomposition and Left inverse [4] have been used in the literature for the inversion of rectangular encoding matrices in SENSE. This paper proposes the use of LU decomposition (for encoding matrix inversion) in SENSE. This work also presents a GPU based parallel implementation of the proposed method to further decrease the computation time. Subjects and Methods: LU decomposition based SENSE reconstruction is implemented on CPU using C-language as well as on GPU using CUDA framework. Figure-1a shows the flow chart of the sequential implementation (CPU based) of the proposed method and Figure-1b shows its parallel implementation (GPU based). In this work, the encoding matrix ‘C’ is first converted into a square matrix ‘A’ by multiplying it with its transpose ‘Ct’. The inverse of matrix ‘A’ is calculated using LU decomposition [2]. ‘A-10 is multiplied by the transpose of encoding matrix ‘Ct’ to get the inverse of the encoding matrix ‘C-1’ (figure-1a).

Discussion/Conclusion: In this work LU decomposition based SENSE is implemented on CPU and GPU. The proposed method successfully reconstructs MR images from the acquired under-sampled k-space data. The results also show that GPU takes less time to compute inversion of the encoding matrix as compared to CPU without any degradation in the quality of the reconstructed images. References: 1) Pruessmann, Klaas P., et al. MRM.1999, 42(5): 952–962. 2) Milan B. Tasic´., et al. IJCM, 2008 85(12): 1865–1878. 3) S.S. Thomas., et al. MRM,.2008 59, 463–468. 4) H. Omer., et al. APMR, 2016. 47(1): 53–61. 5) H. Omer., et al. MAGMA Part A, 2016, 29(1):461–462.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S541

574 Multiframe Denoising in Magnetic Resonance Imaging P. Liebig1, R. Heidemann2, B. Hensel1, D. Porter3 1 Max Schaldach-Stiftungsprofessur fu¨r Biomedizinische Technik, Friedrich-Alexander Universita¨t Erlangen-Nu¨rnberg, Erlangen/ GERMANY, 2HC DI MR TR, Siemens Healthcare, Erlangen/ GERMANY, 3MEVIS, Fraunhofer, Bremen/GERMANY Purpose/Introduction: Denoising methods in Magnetic Resonance Imaging (MRI) are used in combination with Compressed Sensing (CS) [1], with diffusion weighted MRI (dMRI) [2] and in functional MRI (fMRI) [3]. In this abstract we investigate possible benefits of Mutliframe Denoising (MFD) [4, 5] with Haar wavelets for dMRI. This method was already successfully applied to Computed Tomography [4] and Optical Coherence Tomography [5]. Subjects and Methods: A prerequisite for MFD is that multiple images (at least 2) are available that correspond to the number of averages in dMRI. These single images (frames) should have exactly the same image content but only differ in noise. In the following, a short summary of the MFD algorithm is given (compare the flow chart in figure 1). All of these frames undergo a logarithmic transformation followed by a wavelet decomposition. From those transformed frames two sets of weights are calculated.

The significance weights are calculated on the detail coefficients as the mean squared distance of one frame to another. These give an estimate how far the data is corrupted by noise at that position. Correlation weights are calculated for each approximation coefficient by calculating the median of the correlation to each other frame within a small neighborhood. These give a measure if signal is existent at that position. Those two eights were combined as suggested in [5]. Accordingly, the threshold is set to a lower value if there is a high probability that signal is present and noise is low at that location and vice versa. Data was acquired on a MAGNETOM 7T scanner (Siemens Healthineers, Erlangen, Germany) with a 32ch head coil (Nova Medical, Wilmington MA, USA). Results: Figure 2 shows images with no denoising, soft wavelet thresholding and MFD of diffusion weighted data with a maximum of 4 averages.

Discussion/Conclusion: MFD gives in general sharper images and fewer artefacts with even higher SNR. In this study, we only used Wavelet thresholding with Haar Wavelets. In general MFD can be applied with any type of Wavelets and any type of denoising that can be locally controlled by one parameter. However, for MFD, multiple frames are needed and the computation time is about a factor of 10 longer, but it can be easily parallelized. In a next step we will apply MFD in fMRI data sets and also use it in combination with CS. References: [1] Lustig M, Donoho D, Pauly JM. Sparse MRI: The Application of Compressed Sensing for Rapid MR Imaging. Magn Reson Med. 2007, 58 1182–95. [2] Mohana J, Krishnaveni V, GuocY. A survey on the magnetic resonance image denoising methods. Biomed Sig Proc and Contr. 2014, 9 56–59. [3] Pizurica A, Wink AM, Vansteenkiste E, Philips W, Roerdink J. A review of wavelet denoising in MRI and ultrasound brain imaging. Curr Med Imag Rev. 2006, 2 247–60. [4] Borsdorf A, Raupach R, Flohr T, Hornegger J. Wavelet based Noise Reduction in CT-Images using Correlation Analysis. IEEE Trans Med Imaging. 2008.27(12):1685–703. [5] Mayer M, Borsdorf A, Wagner M, Hornegger J, Mardin C, Tornow R. Wavelet denoising of multiframe optical coherence tomography data. Biomed Opt Express. 2012.3(3):572–89. [6] Heidemann RM, Anwander A, Feiweier T, Kno¨sche TR, Turner R. k-space and q-space: combining ultra-high spatial and angular resolution in diffusion imaging using ZOOPPA at 7 T. Neuroimage 2012, 60(2) 967–78.

123

S542

575 Accelerating GRAPPA Operator Gridding (GROG) for L + S GRASP Reconstruction using GPU S. Qazi1, I. Shahzadi2, I. Aslam2, H. Omer2 1 Electrical Engineering, COMSATS Institute of Information Technology, Islamabad/PAKISTAN, 2Electrical Engineering, COMSATS Institute of Information Technology, Islamabad, Islamabad/PAKISTAN Purpose/Introduction: L + S GRASP [1] uses temporal Fast Fourier Transform (TempFFT) and Iterative Soft Thresholding (IST) to reconstruct undersampled radial MRI data. GRAPPA Operator Gridding (GROG) [2] and L + S algorithm is implemented on GPU to accelerate the reconstruction process by exploiting parallelism in GROG, IST and TempFFT. The results show that the proposed implementation of L + S GRASP on GPU significantly improves the image reconstruction time from the acquired undersampled cardiac MRI data. Subjects and Methods:

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 GPU are performed using MATLAB’s parallel computing toolbox on NVIDIA Tesla K40c, with 12 GB memory and 2880 cores. Results: Figure 2(i) and Figure 2(ii) show L + S GRASP reconstruction results of three different phases of DCE liver MRI data using CPU and GPU respectively. The results show that sparse component (S) of L + S reconstruction using GROG and TempFFT provides truly dynamic information without streaking artifacts. Table 1 shows a comparison of the image reconstruction time when the simulations are performed on CPU and GPU. The table shows that the proposed implementation of L + S GRASP with GROG on GPU provides significant gains in the computation time (approximately 12 times for this particular dataset) as compared to the CPU implementation, while maintaining the quality of the reconstructed images.

Discussion/Conclusion: The inherent parallelism of MRI reconstruction algorithms can be exploited by processing independent tasks on GPUs. Parallel implementation of L + S GRASP and GROG on GPU helps to reduce the image reconstruction time with no compromise on the reconstructed image quality. References: [1] R. Otazo et al. MRM V. 73, pp. 1125–1136, (2015). [2] M. Lustig et al. MRM V. 58, pp. 1182–1195, (2008). [3] I. Shahzadi et al. ESMRMB’16, pp S114–S115, (2016). [4] MA. Griswold et al. In:Proc ISMRM, Honululu, Hawaii pp 2410 (2000). [5] N. Seiberlich et al. MRM V. 58, pp 1257–1265 (2007). [6] ‘‘Center for Advanced Imaging Innovation and Research (CAI2R)’’ created by NYU school of Medicine USA (http://cai2r. net/resources/software/grasp-matlab-code). Figure 1 shows conventional L + S GRASP reconstruction pipeline for Golden-Angle Radial sampling using GROG [3]. The receiver coil sensitivity maps are estimated using Walsh method [4] after temporal averaging of the GROG gridded data. Iterative soft-thresholding is used to reconstruct the low-rank (L) and sparse (S) components of the gridded data [4]. GROG decreases the computational complexity of L + S GRASP as it is computationally less intensive as compared to the Non-Uniform FFT (NUFFT) [3, 5]. Due to the presence of parallelism in GROG, IST and TempFFT, this work proposes to execute these independent tasks on GPU to accelerate the reconstruction process. Experiments are performed on the Golden-Angle radial data acquired from 3T scanner at Siemens Healthcare, Erlangen, Germany [6]. The simulations are performed on MATLAB R2014a using Intel Core i74790, 3.60 GHz processor with 16 GB RAM. The simulations on

123

576 Motion Compensated 5D Free-Breathing Whole-Heart Isotropic CINE MRI. An approach based on elastic groupwise registration to promote sparsity R.M. Mencho´n-Lara1, J. Royuela-Del-Val1, A. Godino-Moya1, L. Cordero-Grande2, F. Simmross-Wattenberg1, M. Martı´n-Ferna´ndez1, C. Alberola-Lo´pez1 1 Image Processing Lab, University of Valladolid, Valladolid/SPAIN, 2 Department of Biomedical Engineering, King’s College London, London/UNITED KINGDOM

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Purpose/Introduction: This work proposes a novel under-sampled reconstruction framework for 5D free-breathing whole-heart cardiac CINE MRI. It is based on compressed sensing (CS) principle and applies a groupwise (GW) temporal registration methodology to estimate and to compensate the heart motion. Subjects and Methods: The proposed approach corresponds to the 3D extension of a recently proposed CS reconstruction method for multislice 2D cardiac CINE MRI(1). It is based on a regularized elastic GW registration algorithm, which provides robust motion estimation (in and through-plane) and achieves larger acceleration factors without loss of image quality after reconstruction or, as we show, higher quality for the same acceleration factor. Assuming a golden-radial acquisition(2) and given the respiratory and cardiac motion signals (Fig. 1a), the data are sorted according to the different respiratory and cardiac phases with the approximately same number of spokes in each temporal frame (Fig. 1b). After data binning process, an initial CS reconstruction is performed (Fig. 1c), corresponding to XD-GRASP(3,4). Then, 3D-GW registration is applied over a down-sampled version of this initial reconstruction to estimate the heart motion. Finally, MC-CS reconstruction is obtained with the estimated motion parameters (Fig. 1e). The proposed method has been tested on an isotropic 3D + t cardiac MR scan of a pig. Due to the nature of these animals, respiratory motion is not appreciated in MRI, so a spatiotemporal deformation was synthetically generated to simulate different respiratory positions. Relevant imaging parameters include: voxel size = 1 mm3, field of view = 183 mm3, temporal resolution = 43 ms. Experiments have been performed for different amounts of acquired data with simulated GR.

Results: Figures 2, 3 show the obtained results for two different acceleration factors (AF). XD-GRASP and GW-MC reconstructions are presented for comparison. Better edge delineation, finer details and higher overall quality are appreciated in the case of images reconstructed with the proposed framework, for the same AF.

S543

Discussion/Conclusion: The results indicate that 3D-GW registration achieves a robust estimation of cardiac and respiratory motions, despite the fact that it has been applied over a lower resolution image; aligned images show higher sparsity, which translates into higher quality reconstructions. Robustness is achieved by using all images to compute the transformations as opposed to using a predefined respiratory/cardiac phase as a reference. Future work will focus on the validation of our current results and the inclusion of solutions to avoid the presence of reconstruction artifacts(5). Acknowledgments: The authors acknowledge funds from MINECO and Junta de Castilla y Leo´n through grants TEC2014-57428-R and VA069U16 References: [1] Royuela del Val, J. et al. (2016), Nonrigid groupwise registration for motion estimation and compensation in compressed sensing reconstruction of breath hold cardiac cine MRI. Magn. Reson. Med., 75(4):1525–36. [2] Feng, L. et al. (2016), XD-GRASP: Golden-angle radial MRI with reconstruction of extra motion-state dimensions using compressed sensing. Magn. Reson. Med., 75: 775–788. [3] Feng, L. et al. (2017), 5D whole-heart sparse MRI. Magn. Reson. Med. doi:10.1002/mrm.26745. [4] Royuela-del-Val, J. et al. (2017), Jacobian weighted temporal total variation for motion compensated compressed sensing reconstruction of dynamic MRI. Magn. Reson. Med., 77: 1208–1215. [5] Royuela-del-Val, J. et al. (2017), Jacobian weighted temporal total variation for motion compensated compressed sensing reconstruction of dynamic MRI. Magn. Reson. Med., 77: 1208–1215.

577 L + S Reconstruction of Cardiac MRI with Different Sparsifying Transforms A. Fatima, S. Qazi, H. Omer Electrical Engineering, COMSATS Institute of Information Technology, Islamabad/PAKISTAN Purpose/Introduction: Low-rank plus sparse (L + S) matrix decomposition model was recently proposed to reconstruct dynamic MRI from under-sampled data, which overcomes the limitations of Compressed Sensing (CS) [1]. Sparsity is the number of non-zero pixels present in a dataset. The use of sparsifying transform enforces

123

S544

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

sparsity in L + S model, a necessary condition in CS [2]. In this work, the use of different sparsifying transforms (Temporal Fast Fourier (TempFFT), Wavelet, Contourlet and Discrete Cosine (DCT)) is investigated for sparse recovery in L + S matrix decomposition model. Subjects and Methods: A full rank matrix (M) is decomposed into two components i.e. low rank (L) representing the background of dynamic images and the sparse (S) matrix showing the dynamic information [1]. L + S matrix decomposition solves the following convex optimization problem for image reconstruction [1]:

show that DCT is a better sparsifying transform as it gives low AP, high PSNR and lower RMSE, as compared to TempFFT, Wavelet and Contourlet.

Here T represents the sparsifying transform for S, E is the encoding operator, the regularization parameters for L and S are kL = 0.05 and kS = 0.0001 respectively, which are selected after performing various simulations. In this work, the image reconstruction using different sparsifying transforms (TempFFT, Wavelet, Contourlet, DCT) is presented. Experiments are performed on the fully-sampled cardiac cine data-set acquired from 3T Skyra Siemens scanner at Case Western Reserve University, USA using 30 receiver coils, 11 temporal frames. The acquired data is retrospectively under-sampled for acceleration factor (AF) 6 with variable-density random undersampling. Results: Figure 1(a) shows the results when TempFFT is used as a sparsifying transform in L + S reconstruction, the L and S components are decomposed completely leaving only Sparse in S matrix. In Figure 1 using (b) Wavelet, (d) Contourlet transforms, the sparse information is missing in the S component and the decomposition is not clearly observed. Figure 1(c) shows the blurry images of S component using DCT transform and the separation of components is not clear as well.

Discussion/Conclusion: The results show that TempFFT is a better sparsifying transform as the decomposition of L and S components is more accurate than other transforms and the reconstructed S component is better visible. Computationally simple algorithm and smaller difference in AF, PSNR and RMSE values of TempFFT and DCT makes TempFFT as better sparsifying transform for sparse recovery of Cardiac MRI using L + S matrix decomposition model. References: [1] Otazo et al. MRM V. 73, pp. 1125–1136, (2015). [2] Lustig, et al. EEE 25.2 (2008): 72–82.

The reconstruction results using Artifact Power (AP), Peak Signal-toNoise Ratio (PSNR) and Root-Mean Square Error (RMSE) in Table 1

123

578 Through-Slice Super Resolution Reconstruction of Human Brain MRI N.C. Askin1, A. Klauser1, B. Bejar Haro2, M. Kocher3, F. Lazeyras1 1 Department of Radiology and Medical Informatics, University of Geneva, Geneva/SWITZERLAND, 2School of Computer and Communication Sciences, Ecole Polytechnique Fe´de´rale de Lausanne, Lausanne/SWITZERLAND, 3Biomedical Imaging Group, Ecole Polytechnique Fe´de´rale de Lausanne, Lausanne/ SWITZERLAND Purpose/Introduction: Lengthy MRI acquisitions, such as standard 2D multislice sequence, are prone to motion artifact. A possible strategy to reduce acquisition time and therefore risk of motion is to use multiple stack acquisition technique [1]. The stacks are acquired with low through plane resolution (LR) with slice overlap between stacks [2]. A single high through plane resolution (HR) volume can be reconstructed from the multiple LR stacks. This data acquisition method would be particularly suitable for neonatal MR imaging in clinics due to the subject motions. Subjects and Methods: Data acquisition: One volunteer who provided informed consent, was scanned on a 3T MRI (Prisma-Siemens, Erlangen, Germany) using 64 channel receiver head coil. Turbo Spin Echo sequence with multi band was used to acquire three LR image stacks. LR stacks with through-plane resolution of 4.5 mm were acquired with TR/TE of 5290/154 ms, and 1.0 9 1.0 mm in-plane resolution. Acquisition time for each stack was 43 s. A 1.3 mm through slice shift was applied between each stack (Figure 1). As a

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S545

gold standard, an HR volume was acquired with 1.5 mm slice thickness and the same in-plane resolution (1.0 9 1.0 mm) in 2 min 09 s.

Results: Super Resolution reconstruction of three LR image stacks resulted in a single HR volume. In Figure 3, the image on the left displays the interpolated (spline) first LR image in axial view, the middle shows the Super Resolution reconstructed HR image and the gold standard HR volume is shown on the right.

Data processing: Super resolution reconstruction (SR) method [3] was used to reconstruct an HR volume with the same slice thickness as the gold standard. The framework and equations are shown in Figure 2. The algorithm minimizes the error while reconstructing the HR volume from the limited number of LR stacks with gradient based iterative optimization [4, 5]. Point spread function (PSF) (slice profile) was estimated by a boxcar function.

Discussion/Conclusion: The presented method succeeded in reconstructing an HR volume from multiple LR stacks. SR method could be used in neonatal MR imaging to prevent motion artifacts. In the future work, we aim to reconstruct HR volume with higher number of stacks and more overlap between the stacks. References: [1] Van Reeth, Eric, et al. ‘‘Super resolution in magnetic resonance imaging: A review.’’ Concepts in Magnetic Resonance Part A 40.6 (2012): 306–325. [2] Peeters RR, Kornprobst P, Nikolova M, Sunaert S, Vieville T, Malandain G, et al. ‘‘The use of super-resolution techniques to reduce slice thickness in functional MRI’’. Int J Imaging Syst Technol (2004) 14:131–138. [3] M. Irani and S. Peleg, ‘‘Improving resolution by image registration,’’ CVGIP: Graph. Models Image Process (1991), vol. 53, pp. 231–239. [4] A. Zomet, A. Rav-Acha, and S. Peleg, ‘‘Robust super resolution,’’ in IEEE Proceedings of the Int. Conf. on Computer Vision and Pattern Recognition (CVPR), vol. 1, Dec. 2001, pp. 645–650. [5] P. Vandewalle, S. Su¨sstrunk and M. Vetterli, ‘‘A Frequency Domain Approach to Registration of Aliased Images with Application to Super-resolution’’, EURASIP Journal on Applied Signal Processing (special issue on Super-resolution) (2006), Vol. 2006, pp. Article ID 71459, 14 pages.

123

S546

Interventional MRI

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 gradients. The identified needle position is in very close agreement with the real needle location.

579 Position determination of biopsy needles in interventional MRI using spin echo images with inverted read out gradients J.W. Krug, M. Goerlitz, M. Friebe Department of Electrical Engineering and Information Technology, Otto-von-Guericke-University Magdeburg, Magdeburg/GERMANY Purpose/Introduction: One challenge in interventional magnetic resonance imaging (iMRI) is to control and/or eliminate the image artefacts generated by the interventional used medical devices, e.g. biopsy needles. Magnetic susceptibility artefacts are caused by a local distortion of the MR scanner’s static magnetic field due to the material composition of these devices. Depending on the MR imaging sequence, the static magnetic field strength, and the device’s susceptibility, the resulting artefacts vary in size and shape. Hence, the real needle location, i.e. its tip or centerline, cannot be clearly identified. For MR spin echo (SE) images it is known, that the typical spear shaped artefacts can be altered by changing the direction of the read-out gradient [1, 2]. This work exploits the underlying MRI physics in order to determine the real needle location. Subjects and Methods: A 0.55T benchtop MRI system (magspec, Pure Devices, Wu¨rzburg, Germany) was used to acquire spin echo (SE) images with an echo time (TE) of 4 ms and a repetition time (TR) of 400 ms. Images were taken of a biopsy needle (outer diameter 1.2 mm), placed in a 8 mm diameter cylinder filled with a water copper sulfate solution (1 g/l). The signal amplifications (see Figure 1) in the corners of the artefacts were automatically detected. This step was repeated for the images acquired with inverted read-out gradient. The center of these four points was considered as the real needle location in the axial slice.

Discussion/Conclusion: We developed a method for estimating the position of interventional devices (in our case a biopsy needle) in an MR image using the distinct artefact shape generated by SE images acquired with inverted read out gradients. Although the size of the artefact might seem negligible in the shown SE examples, their size increases with the strength of the MRI’s static magnetic field. In addition, gradient echo (GE) sequences usually used in iMRI generally tend to produce larger artefacts than SE sequences. Hence, the combination of an intermittent SE image acquisition with inverted read gradients for estimating the real or a more precise needle location could improve the precision of MRI guided interventions. References: [1] Ladd, Mark E.; Erhart, Peter; Debatin, Joerg F.; Romanowski, Benjamin J.; Boesiger, Peter; McKinnon, Graeme G.: Biopsy Needle Susceptibility Artifacts. In: MRM 36:646–651 (1991). [2] Schenck, John F.: The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and the second kinds. In: Medical Physics, Vol. 23 (1996). This research was financially supported by the Federal Ministry of Education and Research (BMBF) in context of the ‘INKA’ project (grant number 03IPT7100X).

580 Effect of respiration on the quality of MR thermometry in the head G. Salim1, P. Baron1, D.H.J. Poot1, J.A. Hernandez-Tamames1, M.M. Paulides2, S. Klein3 1 Radiology and Nuclear Medicine, Erasmus Medical Center, Rotterdam/NETHERLANDS, 2Radiation Oncology, Erasmus MC Cancer Institute, ROTTERDAM/NETHERLANDS, 3Biomedical Imaging Group Rotterdam, Departments of Radiology & Nuclear Medicine and of Medical Informatics, Erasmus MC, rotterdam/ NETHERLANDS

Results: For the biopsy needle with an actual outer diameter of 1.2 mm, the artefact had a dimension of 1.6 mm 9 2.2 mm (see Figure 1). Figure 2 shows the two SE images with inverted read out

123

Purpose/Introduction: MR-guided RF hyperthermia requires accurate (DT B 0.5 C) MR temperature measurements during the whole treatment duration in the head & neck region1,2. Although the temperature changes may be measured via the shift in proton resonance frequency3 (PRFS), this technique is sensitive to other sources of temporal magnetic field (B0) changes, like the respiration. In this study, we investigate the errors caused by respiration in temperature measurements in the head region, and we propose a method to reduce them.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Subjects and Methods: The measurements were performed on an interventional 1.5T MRI scanner (GE Optima). A healthy volunteer lay supine in the scanner. First, a 2D multi-echo spoiled gradient echo (ME-SPGR) scan was acquired with: 92 9 92;TE = 1.5, 3.3, 5.1, 6.8, 8.6, 10.3, 12.1, 13.9, 15.6, 17.4 ms; TR = 24.6 ms; FOV = 25 cm; TH = 0.5 cm; BWr = 31.5 kHz; FA = 29; scan duration = 2 s. Second, a 2D spoiled gradient echo scan was acquired dynamically (D-SPGR) during 100 s with: 92 9 92; TE/TR = 11.6/13.6 ms; FOV = 25 cm; TH = 0.5 cm; BWr = 31.5 kHz; FA = 29; one image per second. Both scans were performed during normal respiration. A slice in the axial plane was positioned throughout the nasal cavity and covered the posterior fossa. The MR signal was received with a 22-channels coil. No heating/cooling was applied. Since the volunteer was in steady state conditions, any observed temperature change was assumed to be erroneous4. To determine water and fat proton densities, PDwater and PDfat (Figure 1), the parameters of a multi-peak fat-water model were estimated from data of the ME-SPGR scan. Data from the D-SPGR scan were used to determine the temperature maps via the standard phase-difference technique2. For each time instance, a 1st order regression map of B0 shifts4 (DB0), was created from voxels primarily containing fat (PDfat C 2xPDwater). This allowed us to use the body fat to correct for non-temperature induced phase shifts by subtracting DB0 maps from maps of apparent temperature changes. Results: A region of interest (ROI) of 5 9 5 voxels was selected in the posterior fossa, where the mean temperature changes are plotted w.r.t. time in Figure 2, without and with correction. Over 100 s of normal respiration we observed oscillatory apparent temperature changes of ±7 C at a frequency corresponding to the respiratory cycle. The correction with body fat effectively reduced the temperature errors to within less than ±1 C.

S547

Discussion/Conclusion: Respiration-induced magnetic field changes in the head may cause significant errors in the MR temperature measurements. The order of magnitude is close to that measured by others in the breast5,6. The body fat seems to be a good candidate to reduce these errors. Future work will investigate the potentials of fat referencing in vivo MRT in the head and neck. References: 1. M.M.Paulides, J.F.Bakker, L.W.Hofstetter, et al. Laboratory prototype for experimental validation of MR-guided radiofrequency head and neck hyperthermia Phys. Med. Biol. 59 (2014) 2139–2154. 2. M.M.Paulides, et al. Feasibility of MRI-guided hyperthermia treatment of head and neck cancer, 2014 EUCAP. 3. Y.Ishihara, A.Calderon, H.Watanabe, et al. A Precise and Fast Temperature Mapping Using Water Proton Chemical Shift MRM 1995; 34:814–23. 4. L.W.Hofstetter, D.T.Yeo, W.T.Dixon, et al. Fat-referenced MR thermometry in the breast and prostate using IDEAL, J Magn Reson Imaging. 2012 Sep;36(3):722–32. 5. D.H.J. Poot and S. Klein, Detecting statistically significant differences in quantitative MRI experiments, applied to diffusion tensor imaging, IEEE Trans. Med. Imaging, 2015; 34:5(1164–1176). 6. C.Wyatt, B.Soher, P.Maccarini, et al. Hyperthermia MRI temperature measurement: Evaluation of measurement stabilisation strategies for extremity and breast tumours, Int. J. Hyperthermia, September 2009; 25(6): 422–433. 7. N.H.G.M.Peters, L.W.Bartels, S.M.Sprinkhuizen, et al. Do respiration and cardiac motion induce magnetic field fluctuations in the breast and are there implications for MR thermometry, JMRI.2009 Mar; 29(3):731–5. 8. P.V.Gelderen and C.T.W.Moonen Respiration Induced Changes of Field Homogeneity in the Brain: Implications for fMRI. ISMRM 1998.

123

S548

Musculoskeletal Imaging

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 DM data resulted in a coefficient of repeatability of 3.76 ms for the Tri and of 3.13 ms for the EPG model (Figure 3C).

581 WITHDRAWN 582 Evaluating T2 relaxation times in muscular dystrophy patients; which fitting model to choose, Tri-exponential or Extended Phase Graph? J. Van Asten, L. Heskamp, A. Heerschap Radiology and Nuclear Medicine, Radboud University Medical Center, Nijmegen/NETHERLANDS Purpose/Introduction: MRI of muscles in muscular dystrophy patients reveals fatty infiltration and increased water T2 relaxation times (T2 m)1-4. This increased T2 m is assumed to reflect disease activity5 6 and is therefore an important biomarker to evaluate treatments. Most promising fitting models to determine T2 m are the triexponential (Tri) and extended phase graph (EPG) models6 7 8. As it is important in treatment evaluation to know which of these models is most sensitive to changes in T2 m, we investigated their performance in determining T2 m both healthy and affected muscle. Subjects and Methods: Seven facioscapulohumeral dystrophy patients underwent an MRI of the lower extremity twice, 6 weeks apart. The fourteen multi spin-echo (MSE) images acquired with 17 echoes and echo spacing of 8 ms, were used to voxel-wise fit T2 m with a Tri7 and EPG model6. In addition, for Tri, we investigated the effect of in-homogeneity of the B1 field on the T2 m fit (Figure 1). The muscle with the most oedematous lesion and a region of healthy appearing muscle, were manually delineated and T2 m was determined. The quality of the fit was evaluated by the root-mean-square error (RMSE). Thereafter, T2 m data of five Myotonic Dystrophy (DM) patients, who underwent an MRI twice on 1 day, was used to evaluate repeatability with a Bland–Altman analysis.

Results: EPG fits resulted in a lower average T2 m with a smaller spread than fits by Tri (Figure 2). The T2 m of affected muscle significantly differed from healthy muscle in both models. The increased T2 m was more pronounced for Tri (250%) than for EPG (120%) (Figure 3A). EPG has a significantly lower RMSE than Tri using all 17 echos, while removal of the first 2 echos results in a lower RMSE than that of the EPG (Figure 3B). The Bland–Altman analysis of the

123

Discussion/Conclusion: Although the analysis of the DM data showed a better repeatability for the EPG, the Tri-exponential model reveals a higher differentiation between affected and healthy appearing muscle, which makes Tri more sensitive to detect T2 m changes in longitudinal studies. Leaving out bad B1 voxels does not improve the Tri-exponential fit, but leaving out the first two echos results in a lower RMSE. The EPG fitting model, which inherently compensates for B1 variation, approaches best the T2 relaxation times of muscle, as measured by MR spectroscopy9. References: 1. McNally EM, Pytel P. Muscle diseases: the muscular dystrophies. Annual review of pathology 2007;2:87–109. 2. Janssen BH, Voet NB, Nabuurs CI, Kan HE, de Rooy JW, Geurts AC, et al. Distinct disease phases in muscles of facioscapulohumeral dystrophy patients identified by MR detected fat infiltration. PloS one 2014;9(1):e85416. 3. Kornblum C, Lutterbey G, Bogdanow M, Kesper K, Schild H, Schroder R, et al. Distinct neuromuscular phenotypes in myotonic dystrophy types 1 and 2 : a whole body highfield MRI study. Journal of neurology 2006;253(6):753–61. 4. Wokke BH, Van Den Bergen JC, Hooijmans MT, Verschuuren JJ, Niks EH, Kan HE. T2 relaxation times are increased in Skeletal

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 muscle of DMD but not BMD patients. Muscle & nerve 2016;53(1):38–43. 5. Kim HK, Laor T, Horn PS, Racadio JM, Wong B, Dardzinski BJ. T2 mapping in Duchenne muscular dystrophy: distribution of disease activity and correlation with clinical assessments. Radiology 2010;255(3):899–908. 6. Marty B, Baudin PY, Reyngoudt H, Azzabou N, Araujo EC, Carlier PG, et al. Simultaneous muscle water T2 and fat fraction mapping using transverse relaxometry with stimulated echo compensation. NMR in biomedicine 2016;29(4):431–43. 7. Azzabou N, Loureiro de Sousa P, Caldas E, Carlier PG. Validation of a generic approach to muscle water T2 determination at 3T in fatinfiltrated skeletal muscle. Journal of magnetic resonance imaging : JMRI 2015;41(3):645–53. 8. Lebel RM, Wilman AH. Transverse relaxometry with stimulated echo compensation. Magnetic resonance in medicine 2010;64(4):1005–14. 9. Forbes, Sean et al., Magnetic Resonance Imaging and Spectroscopy. Assessment of Lower Extremity Skeletal Muscles in Boys with Duchenne Muscular Dystrophy: A Multicenter Cross Sectional Study, PLOS ONE 2014;9(9):1–8.

S549 Human in vivo experiments were performed on a 3T GE-MR750 Scanner with an 8-channel transmit-receive knee coil. Parameters are shown in Table 1.

583 View-ordering Schemes for Parallel-Imaging Variable Flip Angle 3D-GRASE in High-Resolution Knee Imaging A. Cristobal-Huerta1, D.H.J. Poot1, M.W. Vogel2, J.A. HernandezTamames1 1 Radiology and Nuclear Medicine Department, Erasmus MC, Rotterdam/NETHERLANDS, 2ASL Scientists Europe, GE Healthcare, Hoevelaken/NETHERLANDS Purpose/Introduction: Specific view-ordering schemes for knee high-resolution images for 3D-GRASE1 acquisitions with k-based parallel-imaging (PI) techniques have not been addressed yet. Viewordering establishes the time order at which each view, or k-space line, is acquired to minimize artefacts and to achieve the desired contrast. View-order optimization is especially important in 3DGRASE because the T2, T2* and phase variations among views for each ET. The purpose of this work is to propose new view-ordering schemes for k-space based PI techniques in 3D-GRASE. These view-ordering schemes will allow to obtain faster scans with reduced specific absorption rate (SAR) than 3D-FSE/TSE in PD-weighted contrast. Subjects and Methods: Two different Cartesian view-ordering schemes were developed combining two available view-ordering techniques to enable parallel imaging in 3D-GRASE: SORT2 encoding strategy combined with linear modulation3 encoding, which splits off-resonance effects and T2-effects in different phase encoding direction and enables to include 2D acceleration while artefacts are mitigated. In SORT-linear modulation encoding strategy, two slightly different approaches are presented. In both, the k-space is filled outwards along ky, where SORT Center-Linear modulation encoding alternates echo trains either ky+ or ky-, while in SORT Center-Out-Linear modulation encoding each echo train equally acquires both. As adjacent echoes in the ET are placed closer in k-space for SORT Center-Linear modulation encoding, the spacing between RF pulses (ESP) is reduced (see Figure 1 A and B). A zero phase ET is acquired for correcting the phase modulation among echoes4. PI was reconstructed by the Autocalibrating Reconstruction for Cartesian sampling algorithm (ARC)5.

Results: 3D-GRASE SORT Center-Linear modulation encoding strategy obtains the shortest ESP. The different view-ordering schemes modulate the magnetization signal in different ways, which is translated into different artefacts: some banding artifacts are observed in the axial and coronal plane in 3D-GRASE (see Figure 2) due to the split of the k-space. SORT Center-Out-Linear modulation also shows some blur in the three orthogonal planes due to longer ESP. Both view-ordering schemes achieve lower specific absorption rate (*40%) and shorter acquisition time (*30%, *3 min).

123

S550

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

584 3D MRI Segmentation of muscle through 2D multi-label propagation at 7T A. Ogier1, A. Foure´1, M. Sdika2, A. Le Troter1, D. Bendahan1 1 CRMBM, UMR CNRS 7339, Aix-Marseille Universite´, Marseille/ FRANCE, 21, UJM-Saint Etienne, CNRS, Inserm, CREATIS UMR 5220, Univ. Lyon, INSA Lyon, Universite´ Claude Bernard Lyon 1, Lyon/FRANCE

Discussion/Conclusion: Image quality in 3D-GRASE strongly depends on the view-order. It is essential to optimize the view-order to achieve the shortest ESP. With an optimal view-order, 3D-GRASE could be an alternative to 3D-FSE/TSE for high-resolution PDweighted knee. References: [1] Koichi Oshio and David A Feinberg. GRASE (gradient and spinecho) imaging: A novel fast MRI technique. Magnetic Resonance in Medicine, 20(2):344–349, 1991. [2] John P. Mugler. Improved three-dimensional GRASE imaging with the sort phase-encoding strategy. Journal of Magnetic Resonance Imaging, 9(4):604–612, 1999. [3] Reed F. Busse, Anja C.S. Brau, Anthony Vu, Charles R. Michelich, Ersin Bayram, Richard Kijowski, Scott B. Reeder, and Howard A. Rowley. Effects of refocusing flip angle modulation and view ordering in 3D Fast Spin Echo. Magnetic Resonance in Medicine, 60(3):640–649, 2008. [4] Jorge Jovicich and David G. Norris. GRASE imaging at 3 Tesla with template interactive phase–encoding. Magnetic Resonance in Medicine, 39(6):970–979, 1998. [5] A. Brau, P. Beatty, S. Skare, R. Bammer. Efficient computation of autocalibrating parallel imaging reconstruction. Proceedings of the 14th Annual Meeting of ISMRM. Seattle, Washington, USA, p. 2462.

123

Purpose/Introduction: Manual segmentation [1] and automated segmentation [2, 3] of individual muscles in magnetic resonance images have been recognized as challenging given the high variability of shapes between muscles and subjects and the discontinuity or lack of visible boundaries between muscles. In the present study, we proposed an original algorithm allowing a 3D semi-automatic segmentation of individual muscles from MR images based on the transversal propagation of manually-drawn 2D masks. Subjects and Methods: Our method is semi-automatic since it requires the manual selection of initial seed regions (composed of a set of 2D labeled masks) corresponding to several slices within the region of interest (3D). It corresponds to the fusion of both propagation and interpolation methods based on several ascending and descending non-linear registrations. Our fusion approach was evaluated on a database consisting of 19 sets of lower leg MR images of healthy volunteers scanned at 7T (T1-weighted images, 200 slices, matrix = 384 9 384, slice thickness = 1.5 mm). The seed regions we used were the manually segmented slices bordering the region of interest and four additional slices located at the borders of lower leg muscles appearing or disappearing along the leg. The DICE similarity coefficient (DSC) and the muscle volume similarity fraction (MVSF) were used to estimate the performance of our method with respect to a manual segmentation of all the muscles of the lower leg. Results: Using only six slices manually-segmented as seed regions for the propagation to the 184 remaining slices as illustrated in Fig. 1, we obtained average DSC values ranging from 0.86 to 0.97 and MVSF values ranging from 0.01 to 0.09. As indicated in Table 1 DSC values higher than 0.90 were obtained for almost all the muscles (GM, SL, SO, FHL, TP, LCLL and ACLL) and the corresponding MVSF values were lower than 0.07. For the FDL muscle, the DSC and MVSF values were 0.86 and 0.09 respectively.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion/Conclusion: We have presented a semi-automated method allowing a segmentation of each lower leg muscles from MR images. The method consisting of an automatic transversal propagation of manually-drawn masks based on several 2D non-linear registration approaches. We mainly showed that our propagated segmentation was very accurate with an averaged DSC value higher than 0.94 and robust for a minimal input of manually-segmented slices (3% of the total slice number).

S551 References: [1] Y. Barnouin et al., ‘‘Manual segmentation of individual muscles of the quadriceps femoris using MRI: a reappraisal’’, JMRI 2014. [2] B. Gilles et al., ‘‘Anatomical modelling of the musculoskeletal system from MRI’’, MICCAI 2006. [3] A. Le Troter et al., ‘‘Volume measurements of individual muscles in human quadriceps femoris using atlas-based segmentation approaches’’, Magma 2016.

123

S552

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Neuroimaging - Clinical 585 Previous antidepressant treatment is associated with increased brain connections in depressed patients: a case control study M.N.T.K. Tran Dong1, R. Colle2, S. Hanadi1, S. Rotenberg2, F. Gressier2, J.-F. Costemale-Lacoste2, A. Rigal2, L. Becquemont3, E. Corruble2, D. Ducreux2 1 IR4 M (UMR8081,CNRS), Saclay University - Paris Sud University, ORSAY/FRANCE, 2Psychiatry, Biceˆtre hospital, Le Kremlin Biceˆtre/ FRANCE, 3Nephrology, Biceˆtre hospital, Le Kremlin Biceˆtre/ FRANCE Purpose/Introduction: Antidepressant drugs (AD) increase hippocampal volume and modify limbic brain activity1,2 in patients with Major depressive episodes (MDE). However, to the best of our knowledge, studies of the impact of AD on brain structural connectome have never been published yet. Thus, we compared brain structural connectome in depressed patients with and without previous AD treatment. Subjects and Methods: Our sample consisted of 55 patients with current MDE (DSM-V). Twenty-three patients had previous AD treatment in the past with a median of 6 months (IQR = 23 months) and 32 had never been treated with AD previously. All of them were antidepressant-free for at least 1 month before the MRI scan, acquired with T1-weighted image and HARDI scans. Brain structural connectome was built based on tractography from 164 cortices and subcortical grey matter structures. These connections were weighted by fiber density and their mean fractional anisotropy (FA). After bivariate analyses, multiple linear regressions were used to compare patients with and without previous AD treatment for connections’ metrics. Statistical significance was ascertained by Student’s t-test with Network Based Statistics, in applying shuffling 10000 times, p-value \ 0.05 and FWER-corrected. Results: Patients with and without previous AD treatment were not different in terms of age, sex and severity as measured by Hamilton Depression Rating Scale (17-items). But patients with previous AD treatment had more frequently recurrent MDE (18 vs 9 patients, p \ 0.05). Brain’s fiber density and FA value did not differ between patients with and without history of MDE in bivariate analysis. Although not different in terms of fiber density neither in bivariate nor in multivariate analysis adjusted for age, sex, smoking status and history of MDE, patients with previous AD treatment had significantly higher FA values in 436 connections (5.6%, mean Cohen’s d = 0.66 ± 0.18) than patients without previous AD treatment (Figure). The difference was nearly symmetric between 2 hemispheres but the fronto-cingulate connections, none of them came from the left anterior and middle cingulate cortices.

123

Discussion/Conclusion: By using structural connectome, we found that patients with previous AD treatment have higher brain’s diffusion capacity without changing its density, almost brain regions were involved as it was observed with functional MRI3,4. This difference is not related to history of MDE and severity of current episode. Together, these findings provided evidence that AD could have longterm effect on brain connections. Subsequent and future researches on imaging-based biomarkers for MDE need to treat carefully this information. References: 1. Fu CHY, Williams SCR, Cleare AJ, et al. Attenuation of the Neural Response to Sad Faces in Major Depressionby Antidepressant Treatment: A Prospective, Event-Related Functional Magnetic Resonance ImagingStudy. Arch Gen Psychiatry. 2004;61(9):877–889. doi:10.1001/archpsyc.61.9.877. 2. Lai C-H, Wu Y-T. Changes in gray matter volume of remitted firstepisode, drug-naı¨ve, panic disorder patients after 6-week antidepressant therapy. J Psychiatr Res. 2013;47(1):122–127. doi: 10.1016/j.jpsychires.2012.09.013. 3. Anand A, Li Y, Wang Y, Gardner K, Lowe MJ. Reciprocal Effects of Antidepressant Treatment on Activity and Connectivity of the Mood Regulating Circuit: An fMRI Study. J Neuropsychiatry Clin Neurosci. 2007;19(3):274–282. doi:10.1176/jnp.2007.19.3.274. 4. Wang L, Xia M, Li K, et al. The effects of antidepressant treatment on resting-state functional brain networks in patients with major depressive disorder. Hum Brain Mapp. 2015;36(2):768–778. doi: 10.1002/hbm.22663.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

586 How accurate are 84 s whole spine scouts from a lumbar MRI in detecting lumbar and, additional thoracic spine fractures? Interdisciplinary preliminary study of patients over 65 years of age

S553 In case of fracture of the LS, sagittal scout showed an overall positive correlation with STIR images. Only one fracture was interpreted as negative in STIR sequence. Inter-reader agreement for sagittal and STIR correlation was almost perfect (Kappa 0.95, 95%CI 0.84–1.0). There was overall good image quality with slight differences between two readers (Figure 3).

M. Kaniewska1, A.J.M. De Beus2, A. Mameghani2, F. Ahlhelm1, R.A. Kubik- Huch1, S.E. Anderson1 1 Radiology, Kantonsspital Baden, Baden/SWITZERLAND, 2 Orthopedic Surgery, Kantonsspital Baden, Baden/SWITZERLAND Purpose/Introduction: Whole spine scouts are performed as part of MRI for labelling of vertebral bodies. To determine if the review of whole spine scout is useful in the setting of MRI of the lumbar spine (LS) for pain review. To examine if additional information can be gained by separate assessment of the thoracic spine (TS). To assess the accuracy of fracture detection in scout and compare with high resolution inversion recovery sequence. Subjects and Methods: Institutional Review Board and informed patient consent was obtained. Patient over 65 years referred for MRI of the LS due to lumbar pain were included. 2 blinded independent readers with experience in spine imaging (Reader 1: radiologist, 3 years’, Reader 2: orthopaedic surgeon, 4 years’) evaluated the whole spine scout (84 s) and compared with the STIR sequence of the LS (3.50 min, control sequence for fracture detection). The TS in the sagittal and coronal whole spine scout was reviewed for further fractures. Inter-reader agreement was measured using Cohen-Kappa statistics. Further fracture characteristics: number of fractures, loss of vertebral body height ([50%), spinal canal stenosis ([30%). Image quality was evaluated (1-poor, 2-medium and 3-good). Results: 108 patients (33 male) were evaluated. On review of sagittal scout, Reader 1 and 2 detected 13 and 11 fractures of the LS, 14 and 15 of the TS, respectively. The inter-reader agreement was substantial for the LS (Kappa = 0.75, 95%CI 0.56–0.96), fair for the TH (Kappa = 0.37 95%CI 0.09–0.66). Coronal scout was not useful in fracture detection. Further fracture characteristics in Figure 1. An example of a sagittal scout with fractures in the LS and TS in Figure 2. If fracture was present in the LS, additional fractures were detected in the TS in 53.85% and 26.67% of patients by Reader 1 and 2, respectively.

123

S554

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: Whole spine sagittal scout may be as accurate in fracture detection as STIR sequence. If the thoracic portion of the whole spine scout is reviewed, additional fractures may be detected, especially if there is a fracture of the LS and the MRI is evaluated by a radiologist. Further imaging of the thoracic spine may greatly assist in patient care. References: 1.Bazzocchi A, Garzillo G, Fuzzi F et al. (2014) Localizer sequences of magnetic resonance imaging accurately identify osteoporotic vertebral fractures. Bone 61:158–163. 2. Bazzocchi A, Spinnato P, Garzillo G et al. (2012) Detection of incidental vertebral fractures in breast imaging: the potential role of MR localisers. Eur Radiol 22:2617–2623. 3. Mai HT, Mitchell SM, Jenkins TJ, Savage JW, Patel AA, Hsu WK (2016) Accessibility of the Cervicothoracic Junction Through an Anterior Approach: An MRI-based Algorithm. Spine (Phila Pa 1976) 41:69–73. 4. Peh WC, Siu TH, Chan JH (1999) Determining the lumbar vertebral segments on magnetic resonance imaging. Spine (Phila Pa 1976) 24:1852–1855. 5. Weiss KL, Storrs JM, Banto RB (2006) Automated Spine Survey Iterative Scan Technique 1. Radiology 239:255–262.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S555

587 Multi-parametric characterization of squamosal cell carcinoma (SCC) of the head and neck using combined 18 F-FDG-PET/MRI

3. Martins EB, Chojniak R, Kowalski LP, Nicolau UR, Lima EN, Bitencourt AG. Diffusion-Weighted MRI in the Assessment of Early Treatment Response in Patients with Squamous-Cell Carcinoma of the Head and Neck: Comparison with Morphological and PET/CT Findings. PLoS One 2015;10:e0140009.

J. Weiss1, K. Zwirner2, D. Thorwarth2, R. Winter2, D. Zips2, C. La Fouge`re3, C. Pfannenberg1, K. Nikolaou1, S. Gatidis1 1 Diagnostic and Interventional Radiology, University of Tuebingen, Tuebingen/GERMANY, 2Radiation Oncology, University of Tuebingen, Tuebingen/GERMANY, 3Nuclear Medicine, University of Tuebingen, Tuebingen/GERMANY

588 Influence of Psychotropic Drugs Consumption on Brain Connectivity in First Episode Psychosis

Purpose/Introduction: SCC of the head and neck is a diagnostically challenging tumor entity [1]. Medical imaging, especially CT and MRI play an essential role for primary staging and therapy monitoring [2]. The recent introduction of combined PET/MR allows for simultaneous assessment of tumor morphology and biological tumor properties using functional MR techniques and metabolic information of FDG-PET [3]. The purpose of this study was to demonstrate the feasibility of voxel-wise tumor characterization in the context of radiotherapy planning. Subjects and Methods: 10 patients (mean age 59 ± 7) with locally advanced primary SCC of the head and neck received multi-parametric combined 18F-FDG-PET/MRI prior to surgery and adjuvant radiotherapy on a clinical 3T PET/MR scanner (Biograph mMR; Siemens Healtineers). All examinations were performed on a radiotherapy planning board with a dedicated planning mask. The protocol consisted of anatomical T2-weighted sequences, functional imaging using diffusion-weighted imaging (DWI) and dynamic contrast-enhanced sequences (DCE) as well as simultaneous acquisition of 18FFDG-PET. SCC were detected on the morphological sequences using 3D regions-of-interest (ROI) with ROI size adapted to the tumor volume. Acquired measurements of DWI (ADCmean and ADCmin), DCE (ktrans and kep) and PET (SUVmean and SUVmax) were correlated for the whole tumor and on voxel-per-voxel basis by an experienced radiologist. Results: Diagnostic image quality was achieved in all patients. Mean tumor volume was 9.5 ± 13 ml. Acquired measurements from DWI, DCE and PET are given as mean ± SD. ADCmean1191 ± 293 10-3 mm2/s; ADCmin360 ± 294 10-3 mm2/s; SUVmean5.3 ± 2.4; SUVtrans 9.8 ± 6.5 min-1; kep 6.1 ± 3.2 min-1. max10.6 ± 5.4/k Correlation analysis revealed a negative correlation for ADC and SUV on a voxel-per-voxel basis (r = 0.46) No correlation was found for the DCE measurements. Discussion/Conclusion: Multi-parametric voxel-wise characterization of squamosal cell carcinoma of the head and neck is feasible with combined 18F-FDG-PET/MRI allowing for simultaneous assessment of tumor morphology and biological tumor properties for more precise therapy planning. References: 1. Kim S, Loevner L, Quon H et al. Diffusion-weighted magnetic resonance imaging for predicting and detecting early response to chemoradiation therapy of squamous cell carcinomas of the head and neck. Clin Cancer Res 2009;15:986–994. 2. Chawla S, Kim S, Dougherty L et al. Pretreatment diffusionweighted and dynamic contrast-enhanced MRI for prediction of local treatment response in squamous cell carcinomas of the head and neck. AJR Am J Roentgenol 2013;200:35–43.

M. Tavares1, S. Reima˜o2, I. Chendo3, R.G. Nunes4 1 Instituto de Biofı´sica e Engenharia Biome´dica, Faculdade de Cieˆncias, Universidade de Lisboa, Lisbon/PORTUGAL, 2 Neurological Imaging Department, Hospital de Santa Maria- Centro Hospitalar Lisboa Norte, Lisbon/PORTUGAL, 3Psychiatry Department, Hospital de Santa Maria- Centro Hospitalar Lisboa Norte, Lisbon/PORTUGAL, 4Institute for Systems and Robotics and Department of Bioengineering, Instituto Superior Te´cnico, Lisbon/ PORTUGAL Purpose/Introduction: A large fraction of patients who suffer a First Episode of Psychosis (FEP) consume psychotropic drugs [1]. Consumption of these stimulants increases dopamine levels in the brain and can cause psychotic symptoms [2]. In this study, we compared diffusion parameters and connectivity metrics in FEP patients with and without consumption habits, to investigate possible brain connectivity changes. Subjects and Methods: Thirty patients with a FEP that had started antipsychotic medication within two weeks of MRI were studied: 19 consumed psychotic drugs (14 only cannabinoids, 5 consumed also other drugs); 11 had no consuming habits. Diffusion-Weighted (DW) images were acquired at 3.0T using a Single Shot Echo Planar Imaging (SS-EPI) sequence: repetition time/effective echo time, 15.6 s/55 ms; axial slices, 80; thickness, 2 mm; field of view, 224 9 240 mm2 (reconstructed pixel size: 0.84 9 0.83 mm2); 32 directions with 1000 s mm-2 b-value; one non-DW image. T1-weighted MPRAGE (Magnetization Prepared Rapid Gradient Echo) images with 0.74 9 0.74 9 1.0 mm3 resolution were also obtained. MIBCA was used for image analysis [3]. Fractional Anisotropy (FA) and Mean Diffusivity (MD) were estimated for all segmented brain regions and network analyses were performed for the whole brain and for a subset of regions associated to drug addiction (caudate, nucleus accumbens, medial and lateral orbito-frontal cortex, and insula) [4]. Characteristic Path Length, Clustering Coefficient and Small Worldness were evaluated. Group comparisons were carried out with Mann–Whitney U tests (P-value \ 0.05 considered significant). Results: MD was decreased in the consuming group in the rightcaudate (0.82 [0.77–0.79] 9 10-3 mm2s-1 versus 0.86 [0.80–1.01] 9 10-3 mm2s-1—median [range]; P-value = 0.0238) and left-nucleus accumbens (0.83 [0.79–0.89] 9 10-3 mm2 s-1 versus 0.87 [0.83–0.92] 9 10-3 mm2 s-1; P = 0.0021). No other differences in diffusion parameters were found. When comparing whole-brain network metrics, only Small Worldness was altered: (5.4 [2.0–7.6] for consuming patients versus 4.4 [2.8–5.7]; P-value = 0.0432)—Figure 1.

123

S556

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [4] Franklin TR, Acton PD, et al. Decreased gray matter concentration in the insular, orbitofrontal, cingulate, and temporal cortices of cocaine patients. Biological Psychiatry 2002; 51:134–42.

589 MR tractography of the brachial plexus I. Ibrahim1, J. Tintera1, A. Skoch1, V. Herynek1, I. Humhej2, M. Hajek1 1 Department of Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague/CZECH REPUBLIC, 2 ´ stı´ nad Labem/ Department of Neurosurgery, Masaryk Hospital, U CZECH REPUBLIC Small-Worldness for whole brain networks for FEP patients with and without consuming habits. Statistical comparison performed using a Mann–Whitney U test. Regarding the reward-specific sub-network, Characteristic Path Length was significantly higher in the consuming group (71.7 [29.3–332.9] mm versus 38.0 [20.6–63.2] mm; P-value = 0.003)— Figure 2.

Characteristic Path Length for FEP patients with and without consuming habits, considering a sub-network for brain regions associated to rewarding. Statistical comparison performed using a Mann–Whitney U test. Discussion/Conclusion: The observed MD changes in the nucleus accumbens and caudate nucleus may be related to microstructural changes in those regions. A possible explanation for the differences in brain network metrics is that the increased Path Length in the reward network may be compensated by an increased connectivity in other areas of the brain, in order to maintain global brain efficiency. In conclusion, DW-MRI may be a useful tool to further evaluate brain changes caused by consumption of psychotic drugs. References: [1] Archie S, Rush BR, Akhtar-Danesh N et al. Substance Use and Abuse in First-Episode Psychosis: Prevalence Before and After Early Intervention. Schizophr Bull 2007; 33 (6): 1354–1363. [2] Gleitman H, Fridlund A, Reisberg D. Psichology. Fundac¸a˜o Calouste Gulbenkian 2011; 9:1044–72. [3] Ribeiro A, Lacerda L, Ferreira H. Multimodal Imaging Brain Connectivity Analysis Toolbox (MIBCA). PeerJ PrePrints 2014;4:1–25.

123

Purpose/Introduction: MR tractography (MRT) and neurography (MRN) are promising methods that can be used to visualize and assess the integrity of the peripheral nerves (PN). These methods are not yet well-established in the PN imaging. We attempt to assess the integrity of brachial plexus structures in control subjects using MRT and MRN techniques (1). Subjects and Methods: Five subjects (3 left-handed and 2 righthanded, mean age 32.8 ± 6.14 years) underwent MR examination on a 3T MR scanner (Siemens Magnetom Trio) using head and neck coils with a following protocol: 1) Diffusion-weighted imaging using a spin-echo echo-planar imaging (SE-EPI) sequence: voxel size 3 9 393 mm3, TR/TE = 7500/ 82 ms, number of diffusion directions 64, two b values: 0, 800 s/mm2. 2) A T2 weighted 3D STIR (short-term inversion recovery) sequence used for high-resolution MR neurography: voxel size 1 9 191 mm3, TE/TR = 149/2000 ms, TI = 160 ms. DTI data were corrected for distortions and eddy current effects using FSL software (2). To reconstruct the fiber pathways, seed regions of interest (ROIs) were manually selected in the EPI image (b = 0 s/ mm2) at different levels of the spinal nerves. The diffusion data were calculated in reconstructed spinal nerves from C5-C7 region using DSI studio (3) and statistically analyzed by two linear mixed-effects models with response variables fraction anisotropy (FA) and mean diffusivity (MD) and explanatory variables ‘laterality’ (categorical) and ‘location’ (C5-C7, modelled as a centered continuous variable and a random effect ‘subject’. The data were Box-cox transformed to improve normality of residuals. Models were stepwise reduced by elimination of insignificant terms. The significance of the factor ‘laterality’ was tested. To correct for multiple comparison (two response variables), computed p-values were multiplied by a factor 2 (4). MR neurogrpahy was calculated in MedINRIA and enhanced with a 3D mode—maximum intensity projection (5). Results: MRT and corresponding MRN are shown in the Figure 1. Both methods enabled reliable graphical reconstruction of the spinal nerves C5-C7 and quantification of diffusion indices. Fractional anisotropy and mean diffusion values are listed in Tab. The linear model showed a significant effect of laterality in the FA values (p = 0.02) and non-significant for MD values, however, statistical results are unreliable due to low number of subjects.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S557 during childhood, affecting approximately 3–6% of school-aged children [1, 2]. In this study, it was aimed to determine the specific changes in sensory motor network which related with ADHD. Subjects and Methods: This study was approved by the local Ethics Committee (Decision no: 2013/689). Anatomical and functional images were collected from participants at rest. Ranging from 9 to 16 years, right handed children with ADHD and typically development (TD) children (n = 20 in each group) participated. fMRI data were preprocessed using Statistical Parametric Mapping (SPM) implemented in MATLAB, and included motion correction, coregistration, spatial normalization to an MNI template, and spatial smoothing. Postprocessing included group independent component analysis (ICA) approach to identify spatially independent and temporally coherent networks. Group ICA was performed using the Group ICA for fMRI Toolbox (GIFT). Group ICA was performed with 30 independent components (ICs) using the Infomax algorithm repeated 50 times with ICASSO which implemented in GIFT. While presented the results, z-score has been selected as 3.5 thresholded. Results: ICs were revealed significantly difference for ADHD and TD groups in one-sample t-test results. After postprocessing IC10 and IC15 was found sensory motor networks. Changed functional activity was observed in sensory motor networks (IC10, IC15) between TD and ADHD groups. Compared to TD group in IC10, it has been shown increased functional activity of bilateral postcentral gyrus, right paracentral lobule, left superior frontal gyrus, bilateral sub-gyral, bilateral middle frontal gyrus; decreased functional activity rigt superior frontal gyrus, left and right precentral gyrus, bilateral medial frontal gyrus in children with ADHD as given Table 1. Additionally compared to TD group in IC15, it has been shown significant increased functional activity of left precuneus and left superior parietal lobule and decreased functional activity of right precuneus in children with ADHD. Functional activity changes in sensory motor areas of brain regions has been presented in Table 1 and Table 2 as Talairach table.

Discussion/Conclusion: MRT and MRN provide unique visualization of the spinal nerves and may be used in pre-surgical planning and post-operative follow-up. Supported by Ministry of Health of the Czech Republic, grant No. 17-28587A and Institutional Support 00023001IKEM. References: 1) Gasparotti R, Lodoli G, Meoded A, Carletti F, Garozzo D, Ferraresi S. Feasibility of diffusion tensor tractography of brachial plexus injuries at 1.5 T. Invest Radiol. 2013 Feb;48(2):104–12.

590 Differences of functional activity in sensory-motor networks at rest in children with attention deficit hyperactivity disorder ¨ zmen2 S¸ . Gengec¸ Benli1, S. Ic¸er1, S. O 1 Engineering Faculty, Biomedical Engineering, Erciyes University, Kayseri/TURKEY, 2Department of Child and Adolescent Mental Health and Diseases, Erciyes University, Kayseri/TURKEY Purpose/Introduction: The correlation patterns in spontaneous neural fluctuations in brain at rest are known as resting-state functional connectivity (RSFC) and could characterize the specific connectivity changes of attention deficit hyperactivity disorder (ADHD). ADHD is the most common neurobehavioral disorder

123

S558

Discussion/Conclusion: Compared to TD group, functional connectivity changes (increased or decreased) in children with ADHD have been demonstrated on motor function areas BAs (4, 6, 8) and somatosensory areas BAs (1, 2, 3, 5, 7). Findings in this study support that somatosensory and motor function problems has been occured in children with ADHD. References: [1] Polanczyk, G., de Lima, M. S., Horta, B. L., Biederman, J., & Rohde, L. A. (2007). The worldwide prevalence of ADHD: a systematic review and metaregression analysis. Am J Psychiatry, 164(6), 942–948. Journal Article. doi:10.1176/ajp.2007.164.6.942. [2] Rohde, L. A., & Halpern, R. (2004). [Recent advances on attention deficit/hyperactivity disorder]. J Pediatr (Rio J), 80(2 Suppl), S61-70. Journal Article. doi:10.1590/S0021-75572004000300009.

591 The white matter structure analysis in patients with fragile X syndrome A. Tulupov1, A. Antonov2, E. Isanova2, A. Savelov3, A. Chupakhin1, D. Yudkin4 1 Laboratory of differential equations, Lavrentyev Institute of Hydrodynamics SB RAS, Novosibirsk/RUSSIAN FEDERATION, 2 Institute of Medicine and Psychology, Novosibirsk State University, Novosibirsk/RUSSIAN FEDERATION, 3Laboratory « MRT TECHNOLOGIES » , The Institute International Tomography Center of the Russian Academy of Sciences, Novosibirsk/RUSSIAN FEDERATION, 4Chromosome pathology group, Institute of Molecular and Cellular Biology SB RAS, Novosibirsk/RUSSIAN FEDERATION Purpose/Introduction: The most common cause of the genetically determined mental retardation is the fragile X chromosome syndrome, a hereditary disease associated with the accumulation of more than 200 CGG repeats in the promoter part of the FMR1 gene. Neurological symptoms include a decrease in the rate of mental development, decreased intelligence, autism spectrum disorders, behavioral aberrations [1]. The purpose: to study the peculiarities of the structural organization of white matter of the brain in patients with the fragile X syndrome by the method of MR-tractography.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Subjects and Methods: The study was performed on the ‘‘Achieva’’ scanner, 1.5T, Philips. The diffusion tensor MRI method with the number of directions of the tensor 15 was used. The obtained data were processed by the MR-tractography method in the FiberTrack application. The following tracts were chosen: the uncinate fascicle (left and right), the inferior longitudinal fascicle (left and right), the cerebral fornix (left and right), the cingulate gyrus (left and right), frontopontine tract (left and right), corpus callosum [2]. The study included a group of 20 children with fragile X syndrome, confirmed by cytogenetic methods, as well as a group of 20 of their mothers. Results: The study was shown a statistically significant decrease in the fractional anisotropy by 0.04 and an increase the diffusion coefficient in the group of patients with the fragile X syndrome compared to the group of premutation for all selected tracts (p \ 0.01), except the cerebral fornix. Thus, the involvement of white matter in the pathological process was shown. More pronounced character of the changes in patients with a complete mutation was shown. The obtained data was shown the involvement of the frontopontine tract in the pathological process, which was not discused previously. Discussion/Conclusion: This study shows that the MR-tractography can be effectively applied for white matter structure analysis in patients with fragile X syndrome and in patients with premutation. The values of the fractional anisotropy and diffusion coefficient were defined. The work was supported by the Russian Science Foundation (the project #17-11-01156 in part of the work on magnetic resonance imaging and the project #15-15-10001 regarding genetic analysis). References: [1] J Grigsby The Fragile X Mental Retardation 1 Gene (FMR1): Historical Perspective, Phenotypes, Mechanism, Pathology, and Epidemiology Clin Neuropsychol 30 (6), 815–833. 2016 Jun 29. [2] Brian P. Hallahan, Michael C. Craig, Fiona Toal, Eileen M. Daly, Caroline J. Moore, Anita Ambikapathy, Dene Robertson, Kieran C. Murphy, Declan G.M. Murphy. In vivo brain anatomy of adult males with Fragile X syndrome: An MRI study NeuroImage; Amsterdam54.1 (Jan 1, 2011): 16–24.

592 Cerebral reorganization after brain metastases treated by stereotactic radiosurgery: multiparametric functional and metabolic 3 Tesla MR approach to assess the long term course S. Wagner1, G. Eichner2, H. Gufler3 1 Neuroradiology, Friedrich-Schiller University, Jena/GERMANY, 2 Mathematical Institute, Justus-Liebig University, Giessen/ GERMANY, 3Radiology, Martin-Luther University Halle-Wittenberg, Halle (Saale)/GERMANY Purpose/Introduction: To describe the long term course of cerebral recovery after stereotactic surgery (SRS) for brain metastases with multiparametric MR-imaging at 3T. Subjects and Methods: Inclusion criterium was a decrease in volume of the lesion on routine follow up scans. Multiparametric MR-examinations at 3T were performed with diffusion- weighted imaging (DWI), dynamic susceptibility perfusion-weighted imaging (DSC) and two-dimensional 1H-MRS examinations. Apparent diffusion coefficient (ADC), regional cerebral blood volume (rCBV), blood flow (rCBF) and mean transit time (rMTT) in the irradiated region as well as phosphorylcholine/glycerophosphorylcholine (Cho), creatine

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 (Cr), N-acetylaspartate (NAA), lactate (Lac), and lipids (Lip), were followed over time. Results: Nineteen lesions in 16 consecutive patients were analyzed. The follow up duration was 9 to 143 weeks (mean 47) with a total of 125 routine follow up scans and 40 multiparametric MR-examinations at 3T. In all lesions Cholin remained higher and NAA was always lower over time compared to the contralateral side. Creatine decreased and was lower after 30 weeks after SKS. If detectable, lesions always showed lactate in the time course. ADC was always higher than the contralateral side. In 75% of the examinations rCBF and rCBV, respectively, was higher in the irradiated region regardless at any time after SRS. Discussion/Conclusion: Knowing the natural course of brain reorganization after SKS for brain metastases may be helpful in interpreting changes in the follow up of patients. References: Will be given in the e-poster.

593 Can Arterial-Spin-Labelling Metrics Serve as ImagingBiomarkers For Isocitrate-Dehydrogenase (IDH)Mutation in Gliomas? A. Alsaedi1, A. Melbourne2, J.-M. Hempel3, X. Golay1, S. Bisdas1 1 Brain Repair & Rehabilitation, Institute of Neurology, University College London, UK, London/UNITED KINGDOM, 2Medical Physics & Biomedical Engineering, Medical Physics & Biomedical Engineering, London/UNITED KINGDOM, 3Department of Neuroradiology, Eberhard Karls University, Tubingen/GERMANY Purpose/Introduction: Recently, many studies have demonstrated the potential of perfusion-MRI in discrimination between tumours types, assessing progression and treatment response, and prediction of the IDH status. However, to our knowledge, there has been no previous study to investigate the value of arterial-spin-labelling (ASL) for differentiating gliomas with and without IDH-mutation. Our aim was to assess IDH mutation status by means of ASL-acquisitions and subsequent histogram-analysis, which takes into account the tumour heterogeneity, rather than any lumped maximum values of tumour blood flow. Subjects and Methods: Forty-four glioma patients, 19 without and 25 with IDH-mutation, were retrospectively included in this study.

S559 Besides the frequently employed mean and maximum absolute-tumour-blood flow (aTBF) and relative-TBF (rTBF) (basal ganglia was used for normalisation), histogram parameters (skewness, kurtosis, and 95-percentiles) were derived from the 2D-PASL cerebral-bloodflow (CBF) maps. The slopes of the cumulative-frequency-curves, as an additional surrogate of tumour heterogeneity, were also calculated. Notably, the TBF metrics were extracted from the whole tumour volumes, as the latter were segmented on FLAIR images. The independent t-test was used to compare the measurements between the tumours with and without IDH-mutation. Results: From the 44-patients, 28 were high-grade and 16 were lowgrade gliomas; the mean-rTBF, 95%-percentile of the proportionaldensity-histogram of the rTBF, cumulative-frequency-slope of the aTBF as well as that of the rTBF were significantly higher in highgrade-gliomas than those of the low-grades. Moreover, 19 of them without and 25 with IDH-mutation (mix of low and high gradegliomas); however, nothing of the extracted parameters can predict the presence of the IDH-mutation. In contrast, by only testing the high-grade-gliomas; max-rTBF was the only estimation that was higher in positive-IDH-gliomas than negative-IDH-gliomas. Bringing glioblastomas-(GBM)into focus, which involved 2 positive-IDHgliomas and 9 negative-IDH-mutation; the mean and max of both aTBF and rTBF were higher in those with IDH-mutation. Discussion/Conclusion: The max-rTBF can be handy for prediction the presence of the IDH-mutation among patients with high-grade gliomas and between those with GBM. Still, this study may provide better results if performed on 3D-ASL (high SNR). References: Just, N. (2014). Improving tumour heterogeneity MRI assessment with histograms. Br J Cancer, 111(12), 2205–2213. http://doi.org/10.1038/bjc.2014.512. Lee S, Choi SH, Ryoo I, et al. Evaluation of the microenvironmental heterogeneity in high-grade gliomas with IDH1/2 gene mutation using histogram analysis of diffusion-weighted imaging and dynamic-susceptibility contrast perfusion imaging. J Neurooncol. 2015;121(1):141-150. doi:10.1007/s11060-014-1614-z.

594 WITHDRAWN

123

S560

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

595 Magnetic resonance perfusion in assessment of the early morphofunctional changes of the central nervous system caused by demyelinating disease 1

2

3

1

A. Tulupov , L. Vasilkiv , E. Isanova , A. Chupakhin 1 Laboratory of differential equations, Lavrentyev Institute of Hydrodynamics SB RAS, Novosibirsk/RUSSIAN FEDERATION, 2 Laboratory « MRT TECHNOLOGIES » , The Institute International Tomography Center of the Russian Academy of Sciences, Novosibirsk/ RUSSIAN FEDERATION, 3Institute of Medicine and Psychology, Novosibirsk State University, Novosibirsk/RUSSIAN FEDERATION Purpose/Introduction: The role of the vascular factor in the pathogenesis of multiple sclerosis is discussed [1]. The structure of the inflammatory foci of the multiple sclerosis is investigated by the most studies and only some studies are aimed at studying the perfusion and hemodynamics of the brain as a whole. The purpose of this study was to assessment of the early morphofunctional changes of the central nervous system caused by demyelinating disease using magnetic resonance perfusion. Subjects and Methods: The study was performed on the ‘‘Achieva’’ scanner, 1.5T, Philips. The MR-perfusion technique with contrast enhanced was used. The CBF, CBV, MTT, TTP were obtained. This study were involved 30 patients: the control group (12 patients) without focal pathology of the brain, the group with multiple sclerosis in the acute stage (4 patients) and the group with multiple sclerosis in the remission stage (14 patients). The McDonald criteria 2010 was used. Results: In the control group, the folowing perfusion parametrs were measured: CBF was 3.015 ± 1.34 ml/100 g/min; CBV was 54.95 ± 21.99 ml/100 g; MTT—18.37 ± 1.21 s; TTP— 17.48 ± 2.24 s. The complex analysis of perfusion parameters in the demyelination foci and visually intact white matter of the brain was performed. At the acute stage in the demyelination foci the significant differences were obtained (p \ 0.01): CBF up to 4.05 ± 1.31; CBV to 72.57 ± 33.81; MTT increase to 17.45 ± 4.07; TTP increase to 17.49 ± 3.69. In the visually intact white matter was detected: CBF up to 3.91 ± 0.59; CBV to 70.83 ± 22.31; However, MTT to 18.11 ± 4.6; TTP to 17.4 ± 4.8 are comparable with control group. The perfusion data in patients with multiple sclerosis at the remission stage was: CBF to 2.02 ± 0.93; CBV to 38.45 ± 18.3 (p \ 0.01); MTT up to 19.3 ± 3.09; TTP to 21.55 ± 10.65; In the white matter was decrease CBF = 2.82 ± 1.14; CBV = 52.89 ± 20.41; MTT up to 19.31 ± 3.3; TTP to 18.7 ± 3.71. Discussion/Conclusion: The obtained data indicates changes in perfusion in the demyelinating foci, depending on the stage of the disease, which allows to use this technique for early diagnosis of multiple sclerosis. The work was supported by the Russian Science Foundation (#17-1101156). References: [1] Ge Y. at all. Dynamic Susceptibility Contrast Perfusion MR Imaging of Multiple Sclerosis Lesions: Characterizing Hemodynamic

123

Impairment and Inflammatory Activity American Journal of Neuroradiology. 2005.

596 Functional magnetic resonance imaging with seedbased analysis as a diagnostic tool for fragile x syndrome A. Tulupov1, Y. Rymareva2, A. Antonov2, E. Isanova2, A. Savelov3, A. Chupakhin1, D. Yudkin4 1 Laboratory of differential equations, Lavrentyev Institute of Hydrodynamics SB RAS, Novosibirsk/RUSSIAN FEDERATION, 2 Institute of Medicine and Psychology, Novosibirsk State University, Novosibirsk/RUSSIAN FEDERATION, 3Laboratory « MRT TECHNOLOGIES » , The Institute International Tomography Center of the Russian Academy of Sciences, Novosibirsk/RUSSIAN FEDERATION, 4Chromosome pathology group, Institute of Molecular and Cellular Biology SB RAS, Novosibirsk/RUSSIAN FEDERATION Purpose/Introduction: Fragile X syndrome (FXS) is the most common inheritable mental disease in children. The main cause of the disorder is CGG•GCC repeat expansion in 5’ promoter region of the FMR1 gene [1]. High disability in terms of mental retardation, reduced quality of life, the presence of medical and psychosocial problems in dealing with these patients, the lack of highly effective diagnostic methods and targeted treatment [2] reflect importance of research of the disorder. Neurobiogical correlates of the cognitive dysfunction haven’t clearly identified. Despite the fact that there are some articles described morphological changes in grey and white matter among these patients [1], [3] these findings don’t reflect functional connectivity among brain regions. Disturbed functional relationships of the regions possibly play a key role of intellectual disability formation. Thus the method of choice for the patients is functional magnetic resonance imaging. Considering cognitive and movement impairments the patients require special resting state fMRI technique. The feature of the study is seed-based analysis performed according to DTI data due to selected models. The purpose of the study is to assess functional connectivity within selected brain regions among patients with fragile X syndrome compare with healthy control group using seed-based analysis of fMRI data processing. Subjects and Methods: The fMRI was performed on ‘‘Achieva’’ (Philips) scanner with a magnetic field strength of 1.5 T. Resting state fMRI with drug-induced sleep was used. Study involved two groups of patients: 17 children with confirmed fragile X syndrome and 8 healthy volunteers. Seed-based correlation analysis were used. Statistical analysis was performed using FSL (fMRI Brain Software Library). Results: Between group analysis (Control/Frax) of ROIs, participated in cingulum bundle formation revealed that patients with fragile x syndrome showed deactivation of left inferior temporal gyrus in moment of activation of the right middle part of cingulum cortex (pvalue 0.0001) and also deactivation of right inferior temporal gyrus in

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 moment of activation of the left middle part of cingulum cortex (pvalue 0.0001), but control group didn’t present the deactivtion. Between group analysis (Control/Frax) of ROIs, participated in inferior longitudinal fasciculus formation confirmed possible role of deactivation in inferior temporal gyrus in face perception impairments leading to social and emotional disturbances of the patients. Actually, the group of Frax showed deactivation in occipital gyruses in the moment of activation of superior temporal gyrus compare with Control group, which didn’t show the pattern (p-value 0.0001). Between group analysis (Control/Frax) of ROIs, participated in uncinate fasciculus formation revealed inadequate deactivation of part of prefronal cortex known as Brodman area 10 during left inferior parietal lobe activation compare with control group, which showed stronger deactivation (p-value 0.0001) and also inadequate (weaker) co-activation the region with right inferior parietal lobe compare with control group, showed stronger co- activation (p-value 0.0001). Discussion/Conclusion: New data of functional status of the brain in patients with fragile X syndrome were received. Seed-based analysis of fMRI data revealed some patterns of functional connectivity for patients with fragile X syndrome. Stronger deactivation in inferior temporal gyrus and particularly within occipital gyruses typically involved in face perception and visual processes respectively, may be a neural correlate of impaired social and emotional interaction with society among the patients. The work was supported by the Russian Science Foundation (the project #17-11-01156 in part of the work on magnetic resonance imaging and the project #15-15-10001 regarding genetic analysis). References: [1] Brian P. Hallahan, Michael C. Craig, Fiona Toal et al. In vivo brain anatomy of adult males with Fragile X syndrome: An MRI study//NeuroImage 54 (2011) 16–24. [2] J Grigsby The Fragile X Mental Retardation 1 Gene (FMR1): Historical Perspective, Phenotypes, Mechanism, Pathology, and Epidemiology//Clin Neuropsychol 30 (6), 815–833. 2016 Jun 29. [3] Hall, Scott S., Robert F. Dougherty, and Allan L. Reiss. ‘‘Profiles of Aberrant White Matter Microstructure in Fragile X Syndrome.’’ NeuroImage: Clinical 11 (2016): 133–138. PMC. Web. 23 May 2017.

597 Acute cerebellitis in children: a series of eight cases

S561 limited but occasionally fulminant. It could be infectious or postinfectious or post-vaccinial. Differentiation of acute infectious cerebellitis from postinfectious cerebellitis is usually based on the presence of fever and absence of a latent period following a nonspecific infection. The neuroimaging findings of cerebellitis are variable and sometimes it can be difficult differentiate from neoplasia. We review the clinical, labarotory, and neuroimaging findings in 8 paediatric cases. Subjects and Methods: We present clinical and radiologic findings, treatment and outcome data of eight children diagnosed with acute cerebellitis in our department between 2010 and 2016. Results: There were 5 girls and 3 boys aged 3–16.5 years, median 8 years. Main presenting symptoms were headache, fever, vomiting and ataxia. Most common MR imaging finding at presentation was bilateral, diffuse signal changes in cerebellar hemispheres. Two patients had hydrocephalus, one had herniation, and one, both hydrocephalus and herniation. Etiological work-up for infectious pathogens revealed mycoplasma pneumoniae and influenza virus in one case each. Two cases underwent lumbar puncture: one was normal, the other showed elevated protein and low glucose. Treatment included steroids in all cases, acyclovir in 7, antibiotics in 2, intravenous immunoglobulins in 1, oseltamivir in one case. No patients required surgery for hydrocephalus or herniation. Four patients had full recovery and four had residual neurologic deficits consisting in cerebellar tremor and ataxia (n = 3) and cognitive deficit (n = 2). Patients with neurological sequalae had cerebellar atrophic changes on follow-up MRI. Patients admitted [7 days after symptom onset recovered with sequalae while those admitted \3 days showed complete recovery. Discussion/Conclusion: There were no fatalities in this series of acute cerebellitis, suggesting early diagnosis and treatment can increase the chance of uneventful recovery. References: 1. De Bruecker Y, Claus F, Demaerel P, Ballaux F, Sciot R, Lagae L, Buyse G, Wilms G. MRI findings in acute cerebellitis. Eur Radiol. 2004 Aug;14(8):1478–83. Epub 2004 Feb 13. Review. 2. Garcı´a-In˜iguez JP, Lo´pez-Piso´n FJ, Madurga Revilla P, Montejo Gan˜a´n I, Domı´nguez Cajal M, Monge Galindo L, Sa´nchez Marco SB, Garcı´a Jime´nez MC. Acute cerebellitis in paediatric patients: Our experience. Neurologia. 2017 Mar 15. pii: S0213–4853(17)30029-4. doi: 10.1016/j.nrl.2017.01.006. [Epub ahead of print] English, Spanish. PubMed PMID: 28318729.

R. Gocmen1, M. Yildirim2, B. Konuskan3, D. Yalnizoglu3, B. Anlar3 1 Radiology, Hacettepe University, Ankara/TURKEY, 2Pediatrics, Hacettepe University, Ankara/TURKEY, 3Pediatric Neurology, Hacettepe University, Ankara/TURKEY Purpose/Introduction: Acute cerebellitis is one of the main causes of acute cerebellar dysfunction in children. Its course is usually self-

123

S562

Neuroimaging - Preclinical 598 Effect of fractionated irradiation on the rat brain: 1H MRS study at 7T

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 In vivo 1H MRS obtained with the SVS technique in the olfactory bulb. On the right is shown a localization of the voxel (3 9 3 9 3mm3) on a morphological T2-weighted reference MR image in the coronal plane.

P. Hnilicova´1, S. Ba´lentova´2, D. Kalenska´3, E. Hajtmanova´4, P. Murı´n4, M. Bittsˇansky´1, D. Dobrota3, J. Lehotsky´1, M. Adamkov2 1 Division of Neurosciences at Biomedical Center Martin, Jessenius Faculty of Medicine in Martin, Comenius University in Bratislava, Martin/SLOVAK REPUBLIC, 2Institute of Histology and Embryology, Jessenius Faculty of Medicine in Martin, Comenius University in Bratislava, Martin/SLOVAK REPUBLIC, 3Department of Biochemistry, Jessenius Faculty of Medicine in Martin, Comenius University in Bratislava, Martin/SLOVAK REPUBLIC, 4Department of Radiotherapy and Oncology, University Hospital Martin, Martin/ SLOVAK REPUBLIC Purpose/Introduction: There are known several radiation-induced brain damages that limit the treatment strategies1,2. Although animal models enable to imitate irradiated injury, it is only poorly explored the early response of the juvenile rat brain to a clinically relevant fractionated irradiation2. The aim of this 1H MRS study was to study the early metabolic post-radiation changes in hippocampus (HI), striatum (S), and olfactory bulb (OB) of the juvenile rats´ brain. Subjects and Methods: Early (18–21 weeks) after the whole-brain irradiation performed by radioactive isotope 60Co (35 Gy in 7 fractions), irradiated (n = 9) and sham-irradiated (n = 5) animals (3 months old Wistar male rats) underwent in vivo 1H MRS on a 7T Bruker BioSpin. For similar brain regions positioning T2-weighted MRI were measured (RARE; TR/TE = 2680/40 ms; NA = 2; FOV = 35 9 35 mm2; 23 slices with 0.5 mm thickness; *2 min). Single-voxel 1H MRS from the OB was obtain within *5 min (PRESS; TR/TE = 1500/20 ms; voxel size = 3 9 3 9 3 mm3; NA = 200). For 1H MRS measurement in HI and S was used 2D CSI (PRESS; TR/TE = 1500/20 ms; NA = 36; matrix = 8 9 8; voxel size = 8 9 10 9 1.6 mm3; FOV = 22 9 22 mm2; *16 min). From selected MR-spectra (OB: Figure 1; HI and S: Figure 2) were quantified NAA + NAAG (tNAA), Cr + PCr (tCr), PCH + GPC (tCho), Gamma-aminobutyric acid (GABA), Glutamate + Glutamine (Glx), and Lactate (Lac). Their ratios to tCr were calculated and statistically evaluated in SPSS (ANOVA).

Position of 1H MRS - CSI grid on the T2-weighted MRI with the FOV of 22 9 22mm2 in the coronal plane. The acquired matrix size was 8 9 8, with a nominal voxel size of 8 9 10 9 1.6 mm3. On the right are spectra from striatum (a) and hippocampus (b). Results: A relative metabolite quantification in the irradiated animals showed significant decrease of tNAA/tCr across all selected regions (HI: p = 0.046; S: p = 0.021; OB: p = 0.040) as well as a significant reduction of GABA/tCr in HI (p = 0.012) and OB (p = 0.009), compare to the control group (Table). Table: Metabolite ratios (mean±SD) in selected brain regions with pvalues expressing differences between irradiated and control group olfactory bulb

tNAA/tCr tCho/tCr GABA/tCr Glx/tCr Lac/tCr

irradiated

control

0.96±0.20 0.29±0.04 0.28±0.15 1.45±0.33 0.20±0.26

1.21±0.16 0.34±0.02 0.54±0.14 1.74±0.16 0.16±0.15

hippocampus pvalue 0.040 0.064 0.009 0.093 0.744

irradiated

control

0.67±0.17 0.22±0.09 0.06±0.03 1.93±0.72 0.16±0.32

1.54±1.18 0.31±0.08 0.13±0.05 2.24±1.64 0.02±0.03

striatum pvalue 0.046 0.126 0.012 0.628 0.360

irradiated

control

0.89±0.54 0.28±0.21 0.08±0.03 2.30±1.63 0.21±0.30

2.30±1.43 0.24±0.15 0.09±0.06 0.95±0.84 0.96±1.20

pvalue 0.021 0.689 0.891 0.114 0.095

Discussion/Conclusion: Clinically relevant fractionated irradiation focused to the juvenile rats´ brains caused significant loss of neuronal function and viability as well as decreased level of the major inhibitory neurotransmitter GABA, which is essential for development and functions of the healthy brain3. In vivo 1H MRS examination of animal irradiation model seems to be useful for modeling and treatment planning strategies. This study was supported by projects: Biomedical Center Martin (ITMS 26220220187), APVV-15-0107 and APVV-14-0088, co-funded from EU sources. References: 1. Greene-Schloesser D, Robbins ME, Peiffer AM, et al. Front Oncol. 2012; 2: 73. 2. Brown RJ, Jun BJ, Cushman JD, et al. Int J Radiat Oncol Biol Phys. 2016; 96: 470–8. 3. Li K, Xu E. Neurosci Bull. 2008; 24: 195–200.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

599 Magnetization transfer ratio in brain bottom areas of non-human primates development

S563 shows how much suppressed by the MTC effect by the change ratio, and let M 0 be the signal strength of the image due to the protons of free water and MS be the signal strength of the image subjected to the MTC effect It is expressed by the equation.

M. Nishio1, Y. Komaki2, F. Seki3, J. Hata4, A. Uematsu3, E. Sasaki5, H. Okano3, A. Furukawa1 1 Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo/JAPAN, 2Live Imaging Center, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN, 3Department of Physiology, School of Medicine Keio University, Tokyo/JAPAN, 4 Brain Science Institute, RIKEN, Saitama/JAPAN, 5Marmoset Research Department, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN Purpose/Introduction: Diffusion Weighted Image (DWI) has been already used for clinical practice and studies such as demyelinating disease typified by multiple sclerosis and myelination with development. For the evaluation of the regions at the base of brain, however, the use of DWI is sometimes difficult because magnetic susceptibility artifacts occurring in echo-planar imaging cause distortion. On the other hand, Magnetization Transfer Ratio (MTR) is good at visualizing brain bottom areas. MTR also reflects myelin of brain, and the image is calculated by using Magnetization Transfer Contrast (MTC) images. In human, myelination progresses rapidly during infancy and continues through the adolescent stage. A number of studies have suggested that MTR value increases in healthy human brain development. However, it remains unclear whether this developmental pattern is comparable across the species. In this study, we assessed age-related changes in magnetization transfer ratio with atlas based analysis especially brain bottom areas in common marmoset. Subjects and Methods: All data were acquired in the total 96 healthy common marmosets (1-24 months old) with 7.0T MRI (Bruker BioSpin). MTC is based on exchange of magnetization between protons that are bound to macromolecules and protons in free water. Quantitative MTR is calculated by using non MT images and MTC images.

Proton in living body bind protons of free water which can freely move and macro molecules such as protein and protons of bound water in which motion is restricted exist. Normally, it is the protons of free water that are detected as signals in MRI, protons of bound water with a short T2 relaxation time are not detected as signals and do not contribute to the image. In the MTC method, protons of bound water are saturated by irradiating an RF pulse (MTC pulse) of the resonance frequency of bound water away from free water. Then, since the protons of free water and the protons of bound water undergo chemical exchange and crossrelaxation, the bound water in the vicinity of the saturated polymer and the protons in the free water exchanges with each other, and as a result, the protons of free water The signal drops. The MTR is an indicator that

MTR image is calculated with MTC image and non MT image. MTR can visualize brain bottom areas. And MTR images were normalized and measured MTR value in a plural region of brain parcels using anatomical atlas label. Results: There are roughly two types of age related MTR change. A type of them is a logarithmic relationship existed between MTR value and age at stria terminalis and fimbria of the hippocampus. In another type, MTR shows a plateau at cranial nerves such as optic tract. The investigated areas were different in rate change of the MTR, indicating the different timing of myelination in each region.

Discussion/Conclusion: MTR can detect the differences of age-related change in brain bottom areas. In the brainstem area that controls

123

S564 life support, it is considered that there is an area that has already completed its development in the fetal stage. On the other hand, because there is also a region that develops after birth, it seems that a difference appears in the rate of change of MTR. These data demonstrate MTR can evaluate the development in brain bottom areas of common marmosets. These results indicate MTR can be a good marker to evaluate the normal brain development of common marmoset, and eventually a suitable parameter to assess braindevelopmental disorders using disease models generated using transgenic techniques. References: 1. Erwin LAB. Quantitative MRI-pathology correlations of brain white matter lesions developing in a non-human primate model of multiple sclerosis. NMR Biomed. 2007; 20:90–103. 2. Gozzi M. A magnetization transfer imaging study of corpus callosum myelination in young children with autism. Biol Psychiatry. 2012; 72:215–220. 3. Engelbrecht V. Age-dependent changes in magnetization transfer contrast of white matter in the pediatric brain. Am J Neuroradiology. 1998; 19:1923–1929. 4. Dubois J. The early development of brain white matter: A review of imaging studies in fetuses, newborns and infants. Neuroscience. 2014; 276:48–71. 5. Fjær S. Magnetization transfer ratio does not correlate to myelin content in the brain in the MOG-EAE mouse model. Neurochem Int. The Authors; 2014;83–84:28–40. 6. Hikishima K. Population-averaged standard template brain atlas for the common marmoset (Callithrix jacchus). Neuroimage. 2011 Feb 14;54(4):2741–9. 7. Hashikawa T. Current models of the marmoset brain. Neurosci Res. 2015 Apr;93:116–27.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 (cerebellum, cerebral cortex), then obtained histograms were fitted with Gaussian distributions using Origin 8.6.0 (OriginLab Corp., USA), and statistical analyses were performed using STATISTICA 12 (StatSoft Inc., USA). Results: In the Figure 1 we can observe example of brain’s crosssection in a horizontal plane for time inversion equal to 500 ms with selected area of cerebellum for control (A) and cuprizone (B) mouse. We can perceive differences in areas of corpus callosum and deep cerebellar nuclei between cuprizone and control mouse, which are associated with demyelinating processes. In the Figure 2 average calculated volume of the myelinated WM and GM are shown in mouse brains areas for selected layer shown in the Fig. 1. Significant difference in myelinated volume [F(2, 12) = 5.3391, p = 0.02195] is observed between control and cuprizone mouse.

600 ‘‘Magnetic resonane imaging of demyelinating processes in the cuprizone model - quantitative analysis in mouse brain ex vivo’’ K. Korga, W. We˛glarz, K. Kalita, K. Jasin´ski, W. Pie˛dzia Department of Magnetic Resonance Imaging, Institute of Nuclear Physics, Polish Academy of Sciences, Krako´w/POLAND Purpose/Introduction: Multiple sclerosis (MS) is an autoimmune, demyelinating disease of the central nervous system (CNS) [1]. MS is characterized by multifocal nerve tissue damage in CNS disseminated at the time. In addition demyelinating process results in damage and disintegration of myelin sheath. It causes disappearance of nerve impulse conduction in damaged cells, leading to significant motor impairment. Diagnosis of disease is based on neurological examination and additional tests, such as magnetic resonance imaging MRI [1, 2]. The range of damage may be examined by using quantitative MRI measuring changes in content of myelin. T1-weighted fast gradient echo pulse sequence MP-RAGE may be used for high-resolution imaging of myelination of the brain [2, 3, 4]. Subjects and Methods: The purpose was to estimate cuprizone-induced demyelination of the entire mice brain ex vivo using MRI. In experiment 16 mouse brains (eight control and eight experimental) were imaged using 9.4T Bruker Biospec 94/20 research MRI scanner, equipped with cryo-coil. The 3D T1-weighted MRI data were collected by sequence MP-RAGE for two values of inversion time (TI equal to 500 and 650 ms), optimized for white matter/gray matter (WM/GM) contrast. Other imaging parameters present below the Fig. 1. The resulting images were analyzed using Fiji distribution of ImageJ (Wayne Rasband, USA) for specific brain regions

123

Discussion/Conclusion: The use of the sequence of MP-RAGE enables analysis of demyelinating processes in the cuprizone model. Analysis of the entire brain volume revealed differences in myelination between healthy and diseased mice in cerebellum, cortex and corpus callosum. The relative loss of myelinated area in cerebellum totals approximately 10%, while in corpus callosum totals approximately 35%. Thereupon demyelinating lesions are milder in cerebellum than in corpus callosum. References: [1] Kipp M., Clarner T., Dang J., Copray S., Beyer C., The cuprizone animal model: new insights into an old story, ActaNeuropathol, 118, 723–736, (2009). [2] Pie˛dzia W., Jasin´ski K., Kalita K., Tomanek B., We˛glarz W. P., White and gray matter contrast enhancement in MR images of the

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S565

mouse brain in vivo using IR UTE with a cryo-coil at 9.4 T, Journal of Neuroscience Methods, 232, 30–35, (2014). [3] Bock N.A., Kocharyan A., Liu J.V., Silva A.C., Visualizing the entire cortical myelination pattern in marmosets with magnetic resonance imaging, Journal of Neuroscience Methods, (185), 15–22, 2009. [4] Pie˛dzia W., Bock N., Jasin´ski K., Kalita K., Stanisz G., We˛glarz W.P., MR Imaging of the Mouse Brain using Cryo-coil at 9.4 T Histology in vivo?, Abstr. of XLV Polish Seminar on Nuclear Magnetic Resonance and Its Applications, Krakow, 1–2 December 2014.

601 Susceptibility artefact determination of pigeon skulls (Columbia livia) D. Flores1, S. Solis2, R. Martin2, F. Vazquez2, O. Marrufo3, A. Rodriguez1 1 Department of Electrical Engineering, UAM Iztapalapa, Mexico City/MEXICO, 2Department of Physics, Faculty of Sciences, UNAM, Mexico City/MEXICO, 3Department of Neuroimage, INNN MVS, Mexico City/MEXICO Purpose/Introduction: The presence of ferrimagnetic crystals in the pigeon’s head (Columba livia) causes susceptibility artefacts in magnetic resonance (MR) images as previously reported [1]. As far as we know, the evaluation of these artefacts has not been investigated in MR images. We used the ASTM Standard F2119-07 to determine the susceptibility artefacts in pigeon’s head images. Subjects and Methods: Images of the head of C. livia were acquired using a Gradient Echo sequence and the parameters: TR/TE = 500/ slice thickness = 1 mm, 5 ms, flip Angle = 200, FOV = 55 9 30 mm2, EXP = 1, matrix size = 256 9 256. All of the experiments were performed on a 7T/21 cm Varian system (Varian, Inc, Palto Alto, CA) equipped with DirectDrive technology and a transceiver 16-rung birdcage coil (16 cm long and a 6.5 cm diameter). We used the software tool SAM (sam-toolbox.sf.net) [2] to analise the susceptibility artefacts in the pigeon’s head images. Images were studied using the automatic procedure to calculate the reference values provided by the image histogram as shown in Fig. 1.

Results: Images were acquired and shown in Fig. 2. The size of the artefacts was measured along the blue and green lines as shown in Fig. 2. With these data, susceptibility artefact size as a function of the position were computed for both lines (blue and green) and also shown in Fig. 2.

Discussion/Conclusion: The maximum size for the blue profile is reached near the frontal. Then, its size decreases as we move away from the forehead towards the tip of the beak. Both profiles depict a local maximum located at the centre of the pigeon’s beak with similar size. This standard test was developed for the evaluation of MR image artefacts from passive implants. We used this standard to study the suceptiblity artefact in pigeon’s heads. This is probably the first attempt to non-invasively study this. Further investigations are required to support these results. References: 1. D. Flores, et. al. Suceptibility effect studies of pigeons with MRI, Magn Reson Mater Phy (2011) 24(Suppl 1): 98. doi: 10.1007/s10334-011-0266-7. 2. F. Gu¨ttler, et. al. Software development for the determination of susceptibility artefacts in MRI after ASTM F2119. Biomed Tech 2012; 57 (Suppl. 1) 2012. doi:10.1515/bmt-2012-4161.

602 Maturational trajectories of cortical brain development in common marmoset F. Seki1, K. Hikishima2, Y. Komaki3, J. Hata4, A. Uematsu1, E. Sasaki5, H. Okano6 1 Department of Physiology, Keio University, School of Medicine, Tokyo/JAPAN, 2Graduate University, Okinawa Institute of Science and Technology, Okinawa/JAPAN, 3Live Imaging Center, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN, 4Brain Science Institute, RIKEN, Saitama/JAPAN, 5Marmoset Research Department, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN, 6Department of Physiology, School of Medicine Keio University, Tokyo/JAPAN Purpose/Introduction: The morphometric studies of cortical brain development in human have uncovered features such as slow volume decline and cortical thining until young adulthood in cerebral cortex1. Typical brain developmental patterns have been emerged, but how these patterns are distinctive features from other animals remain unknown. Comparative analysis with animals makes it possible to reveal this respect. Small non-human primates, common marmosets can be one of the suitable animals for neurodevelopmental studies by maximizing features such earlier maturation (2 years) and cooperative care. We studied postnatal brain development in marmosets focusing on volume as well as thickness changes of cerebral cortex. Subjects and Methods: Using longitudinal sample (23 subjects), the longitudinal MRI was performed to acquire T1-weighted images from 1 to 30 months (adulthood). Individual cerebral cortex image was parcellated into four areas (frontal, parietal, temporal, occipital area). The developmental changes were investigated by measurements of

123

S566 volume and thickness. The growth trajectories were estimated using nonlinear mixed-effects model. Results: The developmental pattern of cerebral cortex by volume measurement was increase-decline. Four regions had similar trajectories but were different in the timing of volume peak and the degree of decline thereafter. The volume peak of occipital area was the earliest (4.2 months) whereas of temporal area was the slowest (7.5 months). The typical developmental pattern by cortical thickness measurement was rapid thining in early childhood, and gradual thining was thereafter which continued until adulthood. Parietal area and frontal area had similar trajectories in volume measurement, but were different in thickness measurement. Whereas it was stabilized after rapid thining in parietal area, gradual cortical thining was observed until adulthood in frontal area. Discussion/Conclusion: The increase-decrease patterns observed by volume measurements indicate that it may be a part of both progressive and regressive developmental events. Given the volume measurement partially reflects synaptogenesis and synaptic pruning as observed in prior studies2, marmosets may share these features. Cortical thining may also reflect synaptic pruning partially3. It is observed both human and marmosets, but the pattern differed in early development. Rapid thining was observed from 1 month in marmosets whereas cortical thicking was reported in early development by human studies. It indicates slower maturation in human than in marmosets brain at birth. These shows a potential that comparative analysis with marmosets in addition to rhesus monkey or mouse could provide an insight into deeper understanding of human brain development and to modeling neurodevelopmental disorders. References: 1. Gogtay N, Giedd JN, Lusk L, et al. Dynamic mapping of human cortical development during childhood through early adulthood. Proc Natl Acad Sci USA. 2004;101(21):8174–8179. doi: 10.1073/pnas.0402680101. 2. Stiles J, Jernigan TL. The basics of brain development. Neuropsychol Rev. 2010;20(4):327–348. doi:10.1007/s11065-010-9148-4 . 3. Shaw P, Kabani NJ, Lerch JP, et al. Neurodevelopmental Trajectories of the Human Cerebral Cortex. J Neurosci. 2008;28(14):3586–3594. doi:10.1523/JNEUROSCI.5309-07.2008.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 isoflurane in oxygen, ketamine and pentobarbital) in T2* and diffusion studies on adult mice. Subjects and Methods: MRI studies were performed in a 7T horizontal magnet: T2* maps (TE = 2.8–41 ms, 10 echoes, TR = 300, 8 Averages, 0.098 mm/pixel), DWI (EPI-segmented, 3 directions, 3 b values: 150, 400 and 1000 s/mm2, TR/TE = 2500/31 ms, 0.19 mm/ pixel) and DTI (EPI-ss, 7 directions, 2 b values: 300 and 1400 s/mm2, TR/TE = 3000/40 ms, 0.19 mm/pixel). Adult CD1 male mice (n = 12) were imaged anesthetized with different combinations: isoflurane in O2 (2%), isoflurane in air (2%), ketamine (75 mg/kg) and pentobarbital (60 mg/kg). After each MRI study, animals were stabled 2 days to recover from anesthesia before the next images set. Five brain regions were analyzed: cortex, prefrontal cortex, thalamus, hypothalamus and hippocampus. Results: Figure 1 resumes the results of the MRI analysis on DTI (Fig. 1C and 1D), DWI (Fig. 1E) and T2* (Fig. 1F) of the different brain zones analyzed (Fig. 1A and 1B).

603 Anesthesia influence in magnetic susceptibility dependent MRI studies D. Calle1, I. Guadilla2, S. Cerda´n Garcı´a-Esteller2, P. Lo´pezLarrubia2 1 Department of Experimental Models of Human Disease, Instituto de Investigaciones Biome´dicas ‘‘Alberto Sols’’-CSIC/UAM, Madrid/ SPAIN, 2Department of Experimental Models of Human Disease, Instituto de Investigaciones Biome´dicas Alberto Sols, Madrid/SPAIN Purpose/Introduction: General anesthesia in preclinical MRI plays an important role, as it is necessary to immobilize the animal and avoid stress to the subject. Although the mechanism of global anesthesia, both inhaled and injected, is not well known, the effects are well studied: a general analgesia, loss of motor reflexes, a state of hypnosis and muscle relaxation are produced in the animal [1, 2]. On these grounds, it is important to study the anesthesia effects on MRI, mainly in susceptibility dependent studies as T2* or fMRI which play an important role in preclinical studies of brain. These sequences are highly influenced by blood flow alterations [3], so anesthesia could alter MRI measurements due to its effects on brain. In this work, we have studied different anesthetic compounds (isoflurane in air,

123

Discussion/Conclusion: The results shows no significant differences between the different anesthetic compounds used, indicating that the election of the anesthesia does not imply changes in the MRI results. References: [1] Miller RD. Anesthesia. 7th ed. Churchill Livingstone, Elsevier; 2010. [2] American Society of Anesthesiologists. Standards: Continuum of Depth of Sedation/Definition of General Anesthesia and Levels of Sedation/analgesia. Available at: [3] Stark DD and Bradley WG. Magnetic resonance imaging. Inc., Nova Iorque, ed. Mosby; 1992.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S567

604 Functional diffusion MRI in an animal model of paclitaxel induced-peripheral neuropathy R.M. Oliveira1, I. Tavares2, P. Lo´pez-Larrubia1 1 Instituto de Investigaciones Biome´dicas, CSIC-UAM, Madrid/ SPAIN, 2Department of Experimental Biology, Faculdade de Medicina do Porto, Porto/PORTUGAL Purpose/Introduction: Chemotherapy-induced peripheral neuropathy (CIPN) is a serious side effect of chemotherapy that frequently determines a reduction or discontinuation of treatment [1]. Paclitaxel is a first-line antineoplastic drug used for several blood and solid tumours. It leads to the occurrence of allodynia and hyperalgesia, and this is often the main reason for reduction or discontinuation of therapy [2]. Using CIPN model, we evaluated the activation of Thalamus and Hypothalamus involved in pain modulation with functional diffusion weighted imaging (fDWI). Subjects and Methods: Animal model. Male adult Wistar rats (n = 6) were induced with chemotherapeutic treatment as described by Polomano et al. [3]. Rats (180–220 g) were treated with paclitaxel during 4 alternated days. The animals were injected intraperitoneally with paclitaxel (2 mg/7 kg) dissolved in 4% Dimethyl Sulfoxide (DMSO). The Imaging experiments were performed in a 7T horizontal magnet (90 mm gradient coil, 360 mT/m maximum intensity gradient and a 38 mm volume resonator) before the paclitaxel induction, and 28 and 56 days after the first day of induction. Diffusion weighted images (9 b values, d = 4 ms, TR = 3000 ms, slices thickness = 1.5 mm) were acquired in three orthogonal directions and processed pixel-by-pixel using a mono- and bi-exponential model [4]. Thalamus and Hypothalamus were manually selected based on the Rat Nervous system atlas [5]. Results: In general, Thalamus and Hypothalamus showed a reduction of apparent diffusion coefficient (ADC) and fast diffusion coefficient (FDC)—from the mono- and biexponential adjustment respectivelyin the control group with the time-course of the pathological state (Fig. 2). Similarly, but in a less visible way, the same seems to occur with the vehicle, contrary to paclitaxel group. On day 28 (T28) no effect of paclitaxel was detected in these areas. At T56 time point, results in hypothalamus showed a higher statistic difference on ADC and FDC (Fig. 1) values comparing with Thalamus. Noteworthy, only in the hypothalamus is observed a statistical reduction of slow diffusion phase (SDP) parameter. No differences were observed over the time in the values of slow diffusion coefficient (SDC) in both areas (Fig. 2).

Discussion/Conclusion: These findings revealed that thalamus and hypothalamus are integrated into pain processing in a late time-course of paclitaxel model induction. The hypothalamus shows a more active role in comparison with the thalamus during CIPN, which is a new finding since most studies do not outstand the role of the hypothalamus in pain transmission/control system. References: [1] Seretny, M., et al., Incidence, prevalence, and predictors of chemotherapy-induced peripheral neuropathy: a systematic review and meta-analysis. PAIN, 2014. 155(12): p. 2461–2470. [2] Campbell, J.N. and R.A. Meyer, Mechanisms of neuropathic pain. Neuron, 2006. 52(1): p. 77–92. [3] Polomano, R.C. and G.J. Bennett, Chemotherapy-evoked Painful Peripheral Neuropathy. Pain Medicine, 2001. 2(1): p. 8–14. [4] Lizarbe, B., et al., Imaging hypothalamic activity using diffusion weighted magnetic resonance imaging in the mouse and human brain. Neuroimage, 2013. 64: p. 448–457. [5] Paxinos, C.W.a.G., The rat brain in stereotaxic coordinates. 6th ed. 2007, Oxford: Elsevier Academic.

605 Difference of the MRI properties and connectome between living and postmortem brain Y. Haga1, J. Hata2, A. Uematsu3, Y. Komaki4, F. Seki3, M. Nishio1, N. Kishi5, E. Sasaki6, H. Okano5, A. Furukawa1 1 Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo/JAPAN, 2Brain Science Institute, RIKEN, Saitama/ JAPAN, 3Department of Physiology, Keio University, School of Medicine, Tokyo/JAPAN, 4Live Imaging Center, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN, 5Department of Physiology, School of Medicine Keio University, Tokyo/JAPAN, 6 Marmoset Research Department, Central Institute for Experimental Animals (CIEA), Kawasaki/JAPAN Purpose/Introduction: Magnetic resonance imaging (MRI) of postmortem brainsis used in physiology and forensic neurology, among other fields of study[1]. It has benefit for brain study because it allows to obtain high resolution and contrast [2].

123

S568 Although ex vivo brains have different MRI properties than in vivo brains due to the chemical changes, fixation time, and so on. Here, we examined the relationships between fixation time and MRI properties using long-term imaging (Study 1), and assessed the differences in a comparison between ex vivo (n = 20) and in vivo (n = 20) brains (Study 2). Subjects and Methods: Common marmosets (Callithrix jacchus) were used in this study. They are have been used in recent neuroscience research [3]. MRI was performed using a 9.4-T BioSpec 94/30 (Bruker Optik GmbH, Ettlingen, Germany) unit and a transmitting and receiving coil with an 86-mm inner diameter (40 mm in the case of the ex vivo brain). We researched the relaxation values, the diffusion values, and connection matrix. Results: Changes in ex vivo brains over long-term fixation (Figure 1). The relaxation values, AD, RD, and MD values decreased immediately after sacrifice. In contrast, FA values showed nearly no change from the in vivo value though the value varied within the margin of error depending on the day.

Comparison between in vivo and ex vivo brains (Figure 2, 3). The T1 relaxation value was significantly different in vivo and ex vivo brains. AD, RD, and MD values were significantly different between in vivo and ex vivo brains. Contrarily, there was no significant difference in FA values and connection matrix between in vivo and ex vivo brains.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: It is supposed that dehydration by the fixation is one of the factors of the decrease of the relaxation time. Therefore, strictly unified perfusion fixation and preservative fixation are required for the study of the ex vivo brain using relaxation values. In addition, It suspected that the decrement of the brain temperature might greatly lower its diffusivity eigenvalues. On the other hand, Furthermore, the FA values and connection matrix were unaffected by tissue fixation and not significantly different between in vivo and ex vivo brains. Therefore, nerve structure assessment using ex vivo brain MRI is likely to be equivalent between ex vivo and in vivo brain MRI. In the future, it is necessary to evaluate the validity of ex vivo brain MRI through comparison of nerve fiber orientation using this technique and other techniques including nerve tracers. References: [1]Seewann A et al., Postmortem verification of MS cortical lesion detection with 3D DIR. Neurology. 2012;78(5):302–308. [2] Robert J. Dawe et al., Postmortem MRI of Human Brain Hemispheres: T2 Relaxation Times during Formaldehyde Fixation. Magn Reson Med. 2009;61(4):810–818. [3] Hikishima, K et al., Atlas of the developing brain of the marmoset monkey constructed using magnetic resonance histology. Neuroscience. 2013;230:102–113.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

New Contrasts 606 Mn(II)-dimercaptosuccinate (Mn-DMSA) in vivo studies as possible tumor-imaging paramagnetic agent W.Y. Ussov1, M.L. Belyanin2, A.I. Bezlepkin1, N.L. Shimanovsky3, V.D. Fillimonov2 1 Lab. of Tomography, Institute of Cardiology, Tomsk/RUSSIAN FEDERATION, 2Organic chemistry and biotechnology, National Research Tomsk Polytechnic University, Tomsk/RUSSIAN FEDERATION, 3Molecular Pharmacology, National Research N.I.Pirogov Medical University, Moscow/RUSSIAN FEDERATION Purpose/Introduction: We attempted to develop a Mn-based paramagnetic contrast agent [1, 2] for imaging of tumor processes, employing avidity of dimercaptosuccinic acid (DMSA) to various neoproliferative structures, and for this aim tested the paramagnetic complex Mn-DMSA in animals with fibrooseous and musculoskeletal tumors. Subjects and Methods: Complex Mn-DMSA was obtained using nanopowder technology from manganese (II) carbonate and dimrcaptosuccinic acid mixed together with addition of trace amounts of water, and then dissolved in buffered saline, delivering finally the 0.5 M solution of Mn-DMSA at pH = 6.5. The Mn-DMSA was injected intravenously slowly to sleeping animals (5 cats and 6 dogs, all veterinary patients, with various musculoskeletal tumors), as much as 0.05 ml of 0.5 M solution per kg of BW. MRI scanning in T1weighted spin-echo mode has been carried out with TR = 500 ms and TE = 15 ms, in sagittal, axial and frontal slices as thin as 2–2.5 mm, to the matrix 256 9 256, with the field of view not larger than 200 9 200 mm. The uptake was reported visually as change in intensity of T1-w MRI scans and quantitatively, with calculating for the T1-w MRI images the index of enhancement (IE) of intensity per voxel, as: IE = (MeanInt of T1-w.MRI)Mn-DMSA/(MeanInt of T1w.MRI)pre-contrast. Also uptake kinetics to the tumor was quantified from blood samples and dynamic uptake for 15 min with 30 s T1weighted SE frames, using Gjedde-Rutland-Patlak plots. Results: The R1 relaxivity of Mn-DMSA was obtained as 3.21/(s*mM). Visually on whole-body T1-w. SE scans the Mn-DMSA induced increase of intensity of the tumor, most prominent in the peripheral areas of tumor and less intense over the central non-perfused areas (fig. 1, demonstrating Mn-DMSA enhancement of nasal fibroosseous sarcoma in a cat).

The values of the IE were over 1.29 (IE = 1.38 ± 0.12) for all peripheral parts of tumors, whereas only 1.15–1.19 in central regiones. The normal musculoskeletal tissues did not enhance after

S569 Mn-DMSA injection, with IE = 1.04 ± 0.02. The uptake kblood-tissue constants were as little as \0.12 ml/min/100 g for non-tumor tissues and over 0.24 ml/min/100 g in vascularised regions of tumors, with slow back diffusion and ktissue-blood \ 0.04 ml/min/100 g. Discussion/Conclusion: Paramagnetic complex Mn-DMSA provides obvious intense enhancement of musculoskeletal tumors in T1-SE MRI, demonstrates intermediate relaxivity R1, non-toxic and makes possible to image tumor processes in animals. The Mn-DMSA complex can be suggested as basic molecule for design of Mn paramagnetic agent for tumor imaging. References: 1. Belyanin M.L., Stepanova E.V., Filimonov V.D., Valiev R.R., ˚ gren H., Borodin O.Y, Ussov W.Y. Design, synthesis and evaluation A of a new Mn contrast agent for MR imaging of myocardium based on DTPA-Phenylpentadecanoic acid complex. Chemical Physics Letters. 2016. V. 665: 111–116. 2. Ussov W.Y., Belyanin M.L., Bezlepkin A.I., Borodin O.Y., Bobrikova E.E., Shimanovskii N.L. Magnetic resonance imaging of brain tumors in dogs using paramagnetic contrast enhancement with Mn(II)-DCTA. Bulletin of Experimental Biology and Medicine. 2016. V.161 (5):661–665.

607 High relaxivity macrocyclic paramagnetic agents for MRI A. Fringuello Mingo1, S. Colombo Serra1, S. Baroni2, C. Cabella1, R. Napolitano1, I. Hawala2, L. Lattuada1, F. Tedoldi1, S. Aime3 1 GI&TO, Bracco Imaging Spa, Colleretto Giacosa/ITALY, 2 Dipartimento di Biotecnologie Molecolari e Scienze per la salute, Universita` degli Studi di Torino, Torino/ITALY, 3Dipartimento di Chimica I.F.M., Universita` degli Studi di Torino, Torino/ITALY Purpose/Introduction: The already noticeable diagnostic capability of Magnetic Resonance Imaging (MRI) was further improved by the development of paramagnetic agents, to such an extent that the contrast-enhanced procedures are nowadays more than one-third of all MRI exams. The commercially available macrocyclic contrast agents (CAs) are well appreciated for their inertness and chemical stability, but are characterized by relaxivity (r1) values[1] lower than certain linear agents and thus by a reduced diagnostic potential. This study aims at proposing two high relaxivity macrocyclic Gd-based CAs (GdCAs) as potential candidates for new optimized MRI products, able to join the properties of high inertness and high efficacy. Two Gd complexes were designed and synthesized with a different strategy in terms of chelating cage (HP-DO3A (1) vs HP-DO2P (2)), keeping the properties of a cyclic structure. Subjects and Methods: A deep relaxometric characterization was carried out, consisting in (i) the acquisition of Nuclear Magnetic Resonance Dispersion (NMRD) profiles at 37 C of 1 mM Gd solution in saline, in saline added with 35 g/L of Human Serum Albumin (HSA), in human plasma (HP) and in a ionized simulated body fluid (i-SBF, that mimics the same ionic content of HP); (ii) the measurement of r1 in saline, in a viscous simulated body fluid (v-SBF, that mimics the same viscosity of HP) and in HP at 37 C and 0.47 T; (iii) the study of longitudinal relaxation rate (R1 = 1/T1) as a function of HSA concentration at 37 C of 0.1 mM Gd solution in saline; (iv) the study of R1 as a function of pH at 37 C of 1 mM Gd solution in saline and in i-SBF. Results: Relaxivity values in HP at 20 MHz/37 C of 9.09 ± 0.25 mM-1s-1 for complex 1 and 10.2 ± 0.15 mM-1s-1 for complex 2 were reached. Obtained values, about two times higher than commercial products of the same category, were detailed

123

S570 interpreted as the result of different additional contributions (protein binding, prototropic exchange, motion slowing at increased viscosity) according to the method discussed in [2]. Discussion/Conclusion: The two proposed GdCAs share high r1 in plasma at clinically relevant magnetic field strength (thanks to moderate albumin binding and a not negligible contribution of mobile protons) and they can thus be considered as potential novel tools for effectively enhancing contrast in MRI images. References: [1] Rohrer et al. Investigative Radiology 2005;40:11. [2] Fringuello Mingo et al. Magn. Reson. Med. doi 10.1002/mrm.26519.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 agar powder concentration of 0.65% and 0.85%, respectively (Figure 1(a-b)). Expected wave speeds are around 3 m/s for simulation phantom, phantom 0.65%, and slightly greater for phantom 0.85%. Additionally, OSS-SNR values with respect to excitation frequency are depicted in Figure 1(c-d). Wave speed maps obtained by k-MDEV inversion are demonstrated in Figure 2.

608 Use of Shear Wave Mode Data in Elasticity Inversion in MR Elastography C. Ariyurek1, S. Ozdemir1, B. Tasdelen2, A.S. Ergun3, Y.Z. Ider2, E. Atalar1 1 National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara/TURKEY, 2Department of Electrical and Electronics Engineering, Bilkent University, Ankara/TURKEY, 3 Department of Electrical and Electronics Engineering, TOBBUniversity of Economics and Technology, Ankara/TURKEY Purpose/Introduction: Previously, it was demonstrated by 3D human brain simulations [1] and phantom experiments [2] that observing high shear wave displacements at mode frequencies during MR elastography (MRE) is possible. In a recent study [3], it was discussed that reconstruction of elasticity at mode frequency may lead to errors. On the contrary, because the amplitude of the shear wave displacement is finite due to tissue damping but significantly higher compared to the ones in non-mode frequencies, including mode frequency data increases strain SNR and elasticity inversion is more accurate, which has been demonstrated in this study using both simulations and experiments. Subjects and Methods: Simulation on a 3D homogeneous phantom was performed using COMSOL Multiphysics (COMSOL, Sweden). Young’s modulus, Poisson’s ratio, density parameters were assigned as 25 kPa, 0.499, 1040 kg/m3 for the phantom, respectively. Damping was included using Rayleigh damping. Motion was induced by rotating the phantom and excitation frequency was swept. Similarly, two homogeneous phantoms were prepared with agar–agar powder concentrations of 0.65% and 0.85%, having similar stiffness values with the simulation phantom. Experiments were conducted in a 3T Siemens Tim Trio MRI scanner with an actuator system similar to head cradle. A SE-EPI pulse sequence with motion encoding gradients was used. The mode frequency of each phantom was determined using the methodology in [2]. Phase images were obtained at 9 different excitation frequencies, including the mode frequency (Figure 1d). To evaluate the performance of inversion, octahedral shear strain SNR (OSS-SNR) [4] and k-MDEV [5] were implemented. Results: Mode frequencies were found by determining the frequencies corresponding to peak normalized displacements, which are 30, 28, and 36 Hz for the simulation phantom, and phantoms with agar–

123

Discussion/Conclusion: As seen in Figure 1, OSS-SNR values are maximum at mode frequencies. Wave speed maps become more accurate and homogeneous when the mode frequency data is included in the inversion (Figure 2), due to high shear wave displacement at mode frequency. However, the input vibration levels must be controlled in order to stay in the safety limits for displacement [6]. As a future work, these results have to be validated in human brain MRE studies.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 References: 1. Ariyurek C, Ider YZ, Gurler N, Ozdemir S, Emek A, Ergun AS, Atalar E. Modes of Shear Waves in Brain MR Elastography. In Proceedings of the 22nd Annual Meeting of ISMRM, Milan, Italy, 2014. p. 4270. 2. Ariyurek C, Ozdemir S, Ergun AS, Ider YZ, Atalar E. Experimental Validation of High Shear Wave Displacement at Mode Frequencies in MR Elastography. In Proceedings of the 24nd Annual Meeting of ISMRM, Singapore, Singapore, 2016. p. 1960. 3. McGrath, Deirdre M., et al. ‘‘Evaluation of wave delivery methodology for brain MRE: Insights from computational simulations.’’ Magnetic Resonance in Medicine (2016). 4. McGarry MDJ, Van Houten EEW, Perrinez PR, Pattison AJ, WeaverJB, Paulsen KD. An octahedral shear strain-based measure of SNR for3D MR elastography. Phys Med Biol 2011;56:N153–N164. 5. Tzschatzsch H, Guo J, Dittmann F, Hirsch S, Barnhill E, Johrens K, Braun J, Sack I. Tomoelastography by multifrequency wave number recovery from time-harmonic propagating shear waves. Med Image Anal2016;30:1–10. 6. Ehman EC, Rossman PJ, Kruse SA, Sahakian AV, Glaser KJ. Vibration safety limits for magnetic resonance elastography. Phys Med Biol 2008;53:925–935.

609 Susceptibility determination using a portable 0.55T small-bore MRI system J.W. Krug1, M. Goerlitz2, M. Friebe1 1 Department of Electrical Engineering and Information Technology, Otto-von-Guericke-University Magdeburg, Magdeburg/GERMANY, 2 Department of Electrical Engineering and Information Technology, Otto-von-Guericke Universita¨t Magdeburg, Magdeburg/GERMANY Purpose/Introduction: The knowledge about a material’s or device’s magnetic susceptibility can be important for different applications. In magnetic resonance imaging (MRI) or especially in MRI guided interventions, the magnetic susceptibility, e.g. of a biopsy or radio frequency ablation needle, is highly important since it affects the homogeneity of the MRI’s static magnetic field. Such field inhomogeneities can cause severe distortions in the acquired MR images. We propose a method to determine the magnetic susceptibility of different materials using phase contrast MRI data. Subjects and Methods: Gradient echo MR images were acquired from three different biopsy needles using a 0.55T benchtop (10 mm diameter bore) MRI system (magspec, Pure Devices GmbH, Germany) and a clinical 3T MRI scanner system (Magnetom Skyra, Siemens, Germany) with flip angle of 30 and TE = 4 ms or TE = 6 ms. The samples were placed in a water copper sulfate solution (1 g/l). The distortions of the static magnetic field were computed based on phase images according

S571 to [1]. The phase images were obtained from the complex MR data and were unwrapped [2]. The theoretical distortion of the static magnetic field caused by a cylindrical object with given geometrical dimensions and susceptibility was assessed using an analytical description [3]. In order to obtain the sample’s magnetic susceptibility, the measured magnetic field distortions were compared to the distortions computed by an analytical approach. Using a least-square optimization, the susceptibility parameters of the analytical description were optimized such that the difference between the measured and simulated magnetic field distortions was minimized. Results: Figure 1 shows the measured and the simulated, optimized magnetic field distortions caused by one of the biopsy needles. For the different biopsy needles, similar susceptibility values were obtained from the 0.55T and the 3T measurements.

Discussion/Conclusion: This work presents a method to estimate a material’s magnetic susceptibility based on MRI phase contrast measurements. A portable MRI system was utilized which constitutes a cheap, more flexible, and higher resolution alternative to a conventional MRI scanner for conducting such experiments. Since the actual magnetic susceptibility of the investigated biopsy needles was not known, further experiments are required to provide an absolute ground truth as reference and calibration for future material tests. The method itself is not only helpful for MRI applications but can be considered as a general material testing procedure. References: [1] E. Mark Haacke, Robert W. Brown, Michael R. Thompson, Ramesh Venkatesan, ‘‘Magnetic Resonance Imaging—Physical Principles and Sequence Design’’, John Wiley & Sons, 1999. [2] Phase Unwrapping Toolxbox, General Engineering Research Institute (GERI), Liverpool John Moores University. [3] K. M. Luedecke, P. Rschmann, RR. Tischler, ‘‘A Fast Calculation Method for Magnetic Field Inhomogeneity due to an Arbitrary Distribution of Bulk’’, Wiley Periodicals, Inc., 2003. This research was financially supported by the Federal Ministry of Education and Research (BMBF) in context of the ‘INKA’ project (grant number 03IPT7100X).

123

S572

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Novel Hardware and Sequences 610 A Feasibility Study of Simultaneous Transmission Method with Multiple Volume Coil for Enhancing |B1+| Homogeneity H. Song, H.-J. Kim, P. Heo, D. Kim, K.-N. Kim Gachon Advanced Institute for health Sciences and Technology, Gachon University, Incheon/KOREA, REPUBLIC OF Purpose/Introduction: Despite of higher signal-to-noise ratio (SNR) under increased main magnet field (|B0|) strength of magnetic resonance imaging (MRI) system, shortened wavelength makes volume radio frequency (RF) coils hard to play roles as local body transmission (Tx) coils due to their inhomogeneous RF transmission field (|B1+|) induced by destructive interference and high power necessity for enlarged volume (1 * 4). In this study, it is proposed to add a body sized loop array Tx-only coil to an ordinary head sized transmission and reception (Tx/Rx) loop array coil for simultaneous RF transmission to improve the |B1+| inhomogeneity under affordable power at ultra-high frequency (UHF). Subjects and Methods: To see the feasibility of multi transmitting and its utility, three types of coil configuration (a body sized coil, a head sized coil and their combination for the proposed method) were modeled (Fig. 1) with Sim4Life (ZMT, Zurich, Switzerland). Each coil was consisted with 16 individual rectangular surface coils. Two types of phantom were chosen to have objective comparison between three coil configurations, an oil phantom (relative permittivity (er) = 78 and electrical conductivity (r) = 0 S/m) and a human phantom from a virtual family (Female, 24 years old). To evaluate the performance of the proposed coil, an operating frequency was set to 297.2 MHz. For all three types of coil configuration, each port was set to have 22.5 of phase offset for the circular polarization. The |B1+| nonuniformities and V-factors of each coil were then compared to evaluate the proposed simultaneous Tx method.

Results: The |B1+| nonuniformities and V-factors were calculated for each center slice of coronal (yz-plane), sagittal (xz-plane) and axial (xy-plane) numerically for comparison (Table 1). The |B1+| distributions of each coil for an oil phantom (Fig. 2) and an Ella phantom (Fig. 3) indicate that the proposed method performs as predicted. The proposed simultaneous Tx coil showed relatively more homogeneous |B1+| and higher V-factor than the ordinary head coil.

123

Discussion/Conclusion: Despite of increased V-factor than the ordinary head coil, the proposed method showed feasibility of being one of solutions for enhancing the |B1+| homogeneity as well as its scalability at UHF. However, since the |B1+| of body sized RF coil varies significantly as |B0| changes, a deeper and diverse study on this method should be conducted to improve |B1+| uniformity for human imaging. References: 1. C. M. Collins, W Liu, W. Schreiber, Q. X. Yang, and M. B. Smith; Central Brightening Due to Constructive Interference. With, Without, and Despite Dielectric Resonance. Journal of Manetic Resonance Imaging 21:192–196 (2005). 2. T. Niendorf, N. K. Sodickson, G. A. Krombach, and J. SchulzMenger; Toward cardiovascular MRI at 7 T: clinical needs, technical solutions and research promises. Cardiac 20:2806–2816 (2010). 3. J. T. Vaughan, C. J. Snyder, L. J. DelaBarre, P. J. Bolan, J. Tian, L. Bolinger, G. Adriany, P. Andersen, J. Strupp, and K. Ugurbil; Whole-

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Body Imaging at 7T: Preliminary Results. Magnetic Resonance in Medicine 61:244–248 (2009). 4. P. V. Moortele, C. Akgun, G. Adriany, S. Moeller, J. Ritter, C. M. Collins, M. B. Smith, J. T. Vaughan, and K. Ug˘urbil; B1 Destructive Interferences and Spatial Phase Patterns at 7 T with a Head Transceiver Array Coil. Magnetic Resonance in Medicine 54:1503–1518 (2005). Acknowledgement: This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Ministry of Education, Science and Technology of the Republic of Korea (grant number: NRF-2014M3C7033998), and a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare of the Republic of Korea (grant number: HI14C1135).

S573 Results: The vibrations caused by the coldhead were observable at all locations. Using the extracted coldhead activity signal monitoring was possible by analyzing the signal amplitudes and frequencies. The additional vibrations that were caused by the gradient coils had no negative effect on the extracted activity signal. An exemplary signal is shown in Fig. 1.

611 Failure prevention and detection of superconductive MRI cooling systems using vibration sensors J.W. Krug, J. Beyer, A. Illanes, M. Friebe Department of Electrical Engineering and Information Technology, Otto-von-Guericke-University Magdeburg, Magdeburg/GERMANY Purpose/Introduction: Magnetic resonance imaging (MRI) systems are typically operated at static magnetic field strengths above 1T. The magnetic field is generated using an electromagnet that is operated in superconducting state. The magnetic field generating coils are cooled to a temperature of approximately 3 K using liquefied helium [1]. For safe and efficient operation, the cooling system has to work reliably all the times. One important device of the MRI’s cooling circuit is the coldhead, which is used to reliquefy the vaporized helium. The condition and performance of the coldhead is typically not directly monitored or is not accessible for the end user or customer operating the MRI scanner system. Complete failure could lead to loss of the magnetic field and high associated costs. It would also be beneficial to know beforehand when problems may occur rather than to wait for and detect the failure. In this work, we propose such a monitoring system for the MRI coldhead. The developed system is based on a measurement and evaluation of the characteristic vibrations or structure-borne noises of the coldhead. Subjects and Methods: Vibrations caused by the MRI’s coldhead were measured using a low-cost piezoelectric sensor. A dedicated circuit was designed to amplify the analog signal. Before analogdigital conversion (sampling rate: 10 kHz), the amplified signal was filtered using 4th order Butterworth low pass filter with a cut off frequency of 3 kHz in order to prevent aliasing effects. Vibrations were measured directly at the coldhead but also at the helium gaspipes outside the MR scanner room with and without MR imaging. During MR imaging, additional vibrations occurred caused by the gradient coils. Using time-frequency analysis, a signal representing the coldhead activity could be extracted.

Discussion/Conclusion: This work presents a method for monitoring the coldhead of an MRI system using piezoelectric elements as vibration sensors. Due to the transmission of the structure-borne noises along the helium gaspipes, the monitoring can also be achieved from the outside of the MRI scanner room. Future work will investigate if a long-term monitoring can be used for predicting a failure of the coldhead. References: [1] Wilson M. Superconducting magnets. United Kingdom: Clarendon Press, 1983. This research was financially supported by the Federal Ministry of Education and Research (BMBF) in context of the ‘INKA’ project (grant number 03IPT7100X).

612 An actively decoupled, double-tuned, receive only RF coil design Y. Ha, C.-H. Choi, N.J. Shah Institute of Neuroscience and Medicine-4, Forschungszentrum Juelich, Juelich/GERMANY Purpose/Introduction: Two separate loops are often used to dual- or triple-tune an RF coil [1, 2]. Although each loop is tuned to two different resonance frequencies, the sensitivity of the coil is reduced, especially at higher frequency, due to the mutual coupling effect between the two loops [3]. A higher frequency blocking trap can be inserted on the lower frequency coil to minimise the coupling current [1, 2]. However, coil sensitivity at the lower frequency is then reduced due to the insertion of the trap. In this work, we designed dual-tuned

123

S574 23

Na/1H coils combining a blocking trap with a PIN-diode to achieve high sensitivity at both frequencies and low coupling at 1H frequency. Subjects and Methods:

Three different configurations were built. Figure 1a shows coupled coils, which do not include a decoupling unit. Figure 2b shows a passively decoupled coil. An LC parallel blocking trap is inserted in a 23 Na loop. Figure 1c shows an actively decoupled double-tuned coil. An LC blocking trap controlled by PIN-diode is inserted in 23Na coil. When the coil is driven at 23Na frequency the active decoupling unit is disconnected from the 23Na loop. Thus, the sensitivity of the 23Na coil is not decreased. When the coil is driven at 1H frequency, the active decoupling unit is then connected to 23Na coil and this reduces the coupling current at 1H frequency. S-parameters and Q-factors were measured using a network analyser (ZNB, Rohde & Schwarz) while the coil was loaded with a cylindrical phantom (110 mm diameter, 100 mm length) containing 45 mmol/L NaCl and 4.7 mmol/L NiSO-4. Q-factors were then compared to those of isolated single-tuned loops of identical dimensions.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results:

Measured S-parameters (Fig. 2) show that all the loops were well tuned and matched. Furthermore, S21 value between loops of coupled coil at 1H frequency was only -12 dB. However, S21 values of passively and actively decoupled coils were -33 and -28 dB, respectively. Table 1 presents the Q-factors of the three different double-tuned coils and those of the equivalent single-tuned loops at the two frequencies of interest.

Discussion/Conclusion: The design of the proposed actively decoupled double-tuned coil is straightforward and it provides low coupling between the loops compared to the coupled coil. Coil sensitivity of the proposed coil design is superior to the other two double-tuned coils at both frequencies. We now intend to perform imaging experiments together with a transmit only double-tuned coil. References: [1] Magill, A. W., Choi, C.-H., Ha, Y., & Shah, N. J. Design and construction of a triple-tuned RF probe for 23Na/31P/1H using traps. ISMRM. 2016;2135. [2] Alecci, M., Romanzetti, S., Kaffanke, J., Celik, A., Wegener, H. P., & Shah, N. J. (2006). Practical design of a 4 Tesla double-tuned RF surface coil for interleaved 1 H and 23 Na MRI of rat brain. Journal of Magnetic Resonance, 181(2), 203–211. [3] Fitzsimmons, J. R., Beck, B. L., & Ralph Brooker, H. (1993). Double resonant quadrature birdcage. Magnetic resonance in medicine, 30(1), 107–114.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S575

613 Optimizing 2D RF Pulse Design for B0 Field Strength for Reduced-FOV Imaging O.C. Eren1, E.U. Saritas2 1 Department of Electrical and Electronics Engineering, Bilkent University, Ankara/TURKEY, 2Department of Electrical and Electronics Engineering and National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara/TURKEY Purpose/Introduction: Reduced field-of-view (FOV) imaging significantly improves high-resolution diffusion weighted imaging (DWI) of small structures [1–8]. Two-dimensional spatially-selective radiofrequency (2DRF) excitation pulses are widely used for this purpose. However, these pulses are considerably long (*16–20 ms), causing off-resonance-induced signal losses in the images. This work shows that the optimum 2DRF design significantly varies at 1.5T vs. 3T, and investigates the optimum design parameters vs. B0. Subjects and Methods: 2DRF pulses were designed in MATLAB, with echo-planar trajectories in excitation k-space (Fig. 1). All pulses had slab thickness FOVslab = 4 cm, 4 mm slice, Nmax = 16, with time-bandwidth product (TBW) in the slice-select (SS) direction TBWSS = 3. TBWslab (i.e., the sharpness of the reduced-FOV) was chosen as a free variable between 3–12, with 9–29 ms resulting pulse durations.

Off-resonance performances were compared via Bloch simulations, for up to ± 128 Hz off-resonances. As refocusing RF, a 180o RF pulse with 3.2 ms duration was utilized. Due to duration/bandwidth mismatches between the 2DRF and 180o RF pulses, the excitation and refocusing profiles do not overlap entirely in the presence of field offresonance, creating a signal loss. Furthermore, the slab profile deteriorates slightly, causing outer-volume excitation, which ultimately aliases back into the image. These factors were quantified to compare the performances of the 2DRF pulses. To visually compare the 2DRF pulses, a digital brain phantom was generated from a fuzzy brain model [9] with 0.5 9 0.5 9 0.25 mm3 voxel size. Two field maps were generated based on the anatomy of the brain phantom. Results: As TBWslab increases, the 2DRF pulse duration increases almost linearly (Fig. 2a). Importantly, the overall signal (quantified as the signal in slab minus the signal aliased) peaks at different TBWslab values at 1.5T versus 3T, meaning that the optimum 2DRF designs are not the same at different B0 strengths. At 1.5T, the optimum 2DRF pulse has TBWslab = 10 with 19.6 ms pulse duration. At 3T, the optimum parameters are TBWslab = 6 with 15.3 ms pulse duration.

For brain phantom simulations (Fig. 3), at 1.5T the shorter 2DRF pulse caused aliasing along the reduced-FOV direction of the image, whereas TBWslab = 10 pulse with 19.6-ms duration performed better despite its longer duration. At 3T, however, the performance of the TBWslab = 6 pulse with a shorter duration is visibly better, demonstrating the off-resonance robustness of shorter pulses.

Discussion/Conclusion: The field off-resonance problem demands different 2DRF pulse designs at 1.5T vs. 3T. Especially at 3T, offresonance robustness needs to be an important design criterion. This work showed that the 2DRF pulses need to be significantly shortened for improved performance at 3T. References: 1. Saritas EU, Cunningham CH, Lee JH, Han ET, Nishimura DG. DWI of the spinal cord with reduced FOV single-shot EPI. Magn Reson Med, 60:468–473, 2008. 2. Banerjee S, Nishimura DG, Shankaranarayanan A, Saritas EU. Reduced field-of-view DWI with robust fat suppression and unrestricted slice coverage using tilted 2D RF excitation. Magn Reson Med, 76:1668–1676, 2016. 3. Finsterbusch J. High-resolution diffusion tensor imaging with inner field-of-view EPI. J Magn Reson Imaging, 29:987–993, 2009. 4. Wheeler-Kingshott CA, Parker GJ, Symms MR, Hickman SJ, Tofts PS, Miller DH, Barker GJ. ADC mapping of the human optic nerve: increased resolution, coverage, and reliability with CSF-suppressed ZOOM-EPI. Magn Reson Med, 47:24–31, 2002. 5. Wilm BJ, Svensson J, Henning A, Pruessmann KP, Boesiger P, Kollias SS. Reduced field-of-view MRI using outer volume suppression for spinal cord diffusion imaging. Magn Reson Med, 57:625–630, 2007. 6. Heidemann RM, Anwander A, Feiweier T, Kno¨sche TR, Turner R. Turner. k-space and q-space: combining ultra-high spatial and angular

123

S576

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

resolution in diffusion imaging using ZOOPPA at 7 T. NeuroImage, 60:967–978, 2012. 7. Finsterbusch J. Improving the performance of diffusion-weighted inner field-of-view echo-planar imaging based on 2D-selective radiofrequency excitations by tilting the excitation plane. J Magn Reson Imaging, 35:984–992, 2012. 8. Cocosco CA, Kollokian V, Kwan RKS, Pike GB, Evans AC. Brainweb: Online interface to a 3D MRI simulated brain database. NeuroImage, 5:425, 1997.

614 Experimental validation of an ultimate signal-to-noise ratio for a circular coil E. Lopez1, S. Solis1, R. Martin1, F. Vazquez1, O. Marrufo2, A. Rodriguez3 1 Department of Physics, Faculty of Sciences, UNAM, Mexico City/ MEXICO, 2Department of Neuroimage, INNN MVS, Mexico City/ MEXICO, 3Department of Electrical Engineering, UAM Iztapalapa, Mexico City/MEXICO Purpose/Introduction: The magnetic resonance image quality strongly depends on the signal-to-noise ratio (SNR). Schnell et. al. proposed an ultimate SNR model of general surface and body coils [1]. A number of different SNR models have been proposed in the past [2]. These theoretical approaches represent a hypothesis that usually require experimentally verification to study their applicability. We experimentally validated this SNR model for a circular-shaped coil with phantom images at 300 MHz. Subjects and Methods: We numerically solved the SNR formula in eq. (1) for a circular coil with a 2-cm diameter at 300 MHz. This SNR equation and the meaning of its variables are reported in [1]. All numerical computations were done with the commercial software tool Mathematica (Wolfram, Champaign, IL, USA). To experimentally validate the eq. [1], we obtained phantom images on a 7 T/21 cm Varian system equipped with Direct Drive technology (Agilent, Inc., Santa Clara, CA, USA), and using a standard a spin echo sequence. T2-weighted axial images of a 2-cm-sphere phantom filled with distilled water were acquired using a circular coil (2 cm diameter), and matrix TR/TE = 300/5.71 ms, FOV = 30 9 30 mm2, size = 256 9 256, slice thickness = 2 mm, and NEX = 20. The experimental array for all imaging experiments is shown in Fig. 1.b).

Results: A comparison of both theoretical (green) and experimental (blue) roll-offs was computed and shown in Fig. 1.a). Phantom images were acquired as shown in Fig. 1.b). With the phantom image data, the experimental roll-off was computed along the yellow line as indicated in Figure 1.b).

123

Discussion/Conclusion: Eq. (1) was successfully solved for a circular coil using a numerical approach. Within the vicinity of the coil plane, a slightly disagreement can be observed. This remains to be fully investigated to find out the possible source of error. The Fig 1.a comparison shows an exceptional agreement despite the fact that a numerical solution was used. This is an important step forward to support the development of RF coils for high field MRI, because the electromagnetic frame behind eq. (1) represents full wave calculations of the SNR for surface coils. References: 1. Schnell, W., Renz, W., Vester, M., & Ermert, H. (2000). Ultimate signal-to-noise-ratio of surface and body antennas for magnetic resonance imaging. IEEE Trans Antennas Propag, 48(3), 418–428. 2. Pfrommer, A., Henning, A. (2017). On the Contribution of CurlFree Current Patterns to the Ultimate Intrinsic Signal-to-Noise Ratio at Ultra-High Field Strength. NMR Biomed.

615 Theoretical study of SNRSENSE for a slotted coil array S. Solis1, R. Martin1, F. Vazquez1, O. Marrufo2, A. Rodriguez3 1 Department of Physics, Faculty of Sciences, UNAM, Mexico City/ MEXICO, 2Department of Neuroimage, INNN MVS, Mexico City/ MEXICO, 3Department of Electrical Engineering, UAM Iztapalapa, Mexico City/MEXICO Purpose/Introduction: The coil array development for parallel magnetic resonance imaging (pMRI) has been mainly based on the study of noise (g factor). This factor is not very useful on its own to assess coil performance, because the baseline signal-to-noise ratio (SNR) is not considered. We theoretically studied the performance of a slotted coil array for pMRI. The slotted surface coil has proved to outperform the circular coil and it has lower SAR [1]. We studied the SNRSENSE of a 4-slotted coil array for pMRI and compared to a similar circular coil array, see Fig. 1.a. To the best of our knowledge, this is one the first attempts to study coil array design for pMRI following this approach.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S577

Discussion/Conclusion: We derived an analytical SNRexpression to theoretically calculate the performance of this coil array for SENSE MRI. An important theoretical improvement can be appreciate from Fig. 2. These are very encouraging results because the reference point is located at one radius distance, where the coils tend to show low SNR. The slotted coil array outperformed the convencional coil array for distant points. The present coil configuration guarantees that no unwanted interaction occurs. However, the region of interest is not fully covered as usually sought in an standard MRI experiment. It is remained to experimentally verify these theoretical results. References: 1. Solis-Najera, S. E. et. al. (2015) MAGMA, 28, 599. 2. Reykowski, A. (2002). In Proc. 10th Annual Meeting ISMRM (Vol. 905). 3. Hidalgo, S. S. et. al. (2012) J. App. Phys. 112(3), 034901. 4. Roemer, P. B. et. al. (1990) Mag. Res. Med. 16, 1992.

616 Relationship between signal enhancement with dipolar technique and viscoelasticity of tissues containing macromolecules Subjects and Methods: We derived the SNRSENSE expression for the coil array above from [2]: SNRSENSE = SNRslot/gR1/2 (1), where SNRslot represents the SNR slotted coil array, R is the acceleration factor and. S11, S12, S13, S14 represent the signal received by the coils 1, 2, 3, and 4, for point 1, as shown in Figure 1.b. SNRslot was derived for an 8 slot configuration from with b = 20 cm, c = 4 cm, and a = 5 cm mm [3]. This SNR expression of the slotted coil together with eq. (1) were used to theoretical compute the SNRSENSE as a function of position for point 1 (red point in Fig. 1.b). All calculations were done at 300 MHz and using specially-written MATLAB programmes (MathWorks, Natick, MA, USA). Similarly, SNRSENSE calculations were done for a 4-circular coil array with the same dimensions [4]. Results: An analytical expression for the g factor for both the slotted and circular coils were derived from [2]. With this mathematical frame, we calculated the SNRSENSE for both coil arrays. Fig. 2 shows a comparison of the roll-offs for the two coil arrays and R = 2.

E. Mougel, P.M. Lefebvre, K. Tse Ve Koon, D. Grenier CREATIS, Univ. Lyon; CNRS UMR 5220; INSERM U1206; INSALyon; UJM-Saint Etienne; Universite´ Lyon1, Villeurbanne/FRANCE Purpose/Introduction: MR contrast mechanisms are not easy to achieve for some biological tissues such as tendon, cartilage and other short-T2 tissues; the main reason being dipolar interaction between spins inducing rapid vanishing of the signal. Therefore, a sequence based on Redfield’s works1 was developed in order to cancel the signal loss induced by dipolar exchange2 (Magic Echo Pulse Sandwich (MEPS) sequence). This present work studies the relationship between the signal enhancement obtained using a customized T2q sequence and the viscoelasticity parameters of calibrated phantoms and investigates an eventual link between dipolar interaction presence and viscoelastic properties.

123

S578

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Subjects and Methods: Samples of Plastisol were prepared using different concentrations of softener. During the experiment, two acquisitions were performed successively on each plastisol samples. First, the storage (G’) and loss (G’’) moduli were estimated using the Magnetic Resonance Elastography (MRE) technique (3). Acquisition parameters are presented in the table 1. Then, two MR images were acquired using T2q sequence consisting of a Spin-Echo (SE) sequence followed by a MEPS sequence (parameters in table 1). MR images sets were taken at different echo time values (TE) in order to estimate the signal decrease over time. For each sample, cartographies of transverse relaxation times (T2 and T2q) were plotted for SE and MEPS sequences, respectively.

Comparing figure 1a and figure 1b, the signal enhancement is 120%. Figure 2 shows the evolution of T2 et T2q as a function of G’ and G’’. These relaxation times decrease with increasing G’ and G’’. For the chosen range of G’, the T2 values evolve from 45.9 ± 1.8 ms to 39.9 ± 1.7 ms while, T2q values range between 88.7 ± 3.1 ms and 52.0 ± 4.3 ms. The variation range of T2q values is three times greater than that of T2 for both G’ and G’’. Note that the interval of variation of G’’ is only 6 kPa.

Results:

Discussion/Conclusion: T2 and T2q are both strongly correlated to G’ and G’’ values. Indeed T2q values is in our case much more sensitive than T2 ones within the ranges of G’ and G’’ value chosen. Thus T2q values seems to correlate with the storage and loss moduli, in the range of usual biological viscoelasticity parameters4. Properly calibrated, this innovative technique keeps the advantages of fast and simple classic sequence and may give an easy access to viscoelasticity parameters of short-T2 tissues. Acknowledgement: Authors would like to thank LabEX PRIMES (ANR-11-LABX-0063). References: 1. A.G.Redfield, Phys. Rev., doi: 10.1103/PhysRev.98.1787. 2. D. Grenier and al., doi: 10.1006/jmre.2000.2188. 3. P.M. Lefebvre and al., doi: 10.1109/EMBC.2016.7590924. 4. K. Riek and al., doi: 10.1016/j.jbiomech.2010.12.031.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S579

Peaks and Valleys - MR Spectroscopy 617 Comparison of DCE-MRI and hyperpolarized 13CMRSI in cancer canine patients C. Eschen1, P. Holst2, M. Lundemann Jensen1, H. Bo Wiberg Larsson1, A. Kjær1, J. Henrik Ardenkjær-Larsen3, A. T. Kristensen2, A. Espe Hansen1 1 Department of Clinical Physiology, Nuclear Medicine & PET and Cluster for Molecular Imaging, Rigshospitalet, Copenhagen/ DENMARK, 2Department of Veterinary Clinical Sciences, Faculty of Health and Medical Sciences, University of Copenhagen, Frederiksberg C/DENMARK, 3Department of Electrical Engineering, Technical University of Denmark, Kgs. Lyngby/DENMARK Purpose/Introduction: Dual modality imaging with Positron Emission Tomography (PET) and hyperpolarized 13C magnetic resonance spectroscopy imaging (MRSI) has potential in phenotyping of cancer [1]. Hyperpolarized MRSI employs intravenous injection of a hyperpolarized substrate, and could be influenced by tissue perfusion. Here we compare hyperpolarized 13C MRSI with Dynamic Contrast Enhanced MRI (DCE-MRI) perfusion. Subjects and Methods: Three canine cancer patients diagnosed with thyroid carcinoma were examined using PET/MRI (Siemens) and a 1 13 H/ C flex coil (RAPID Biomedical). 18F-FDG-PET was performed 60 min post injection (8 MBq/kg). Hyperpolarized [1-13C]pyruvate was prepared using SpinLab (GE Healthcare) [1] and imaged with 2D CSI [TR 80 ms, flip angle 10, matrix 16 9 16, voxel size 10 9 10 9 13 mm3]. DCE-MRI utilized 3D VIBE (flip angle 14, TR 2.94 ms, TE 0.91 ms, time resolution 2.6 s). The flow F was estimated using deconvolution with Tikhonov regularization. A 2-compartment model [2] yielded blood volume Vb and permeability Ki. All data in tumor ROI’s were resampled to CSI voxels, pooled and z-normalized. A voxelwise correlation of 13C metabolites (lactate, pyruvate, total 13C and lactate to total 13C ratio) and DCE perfusion (Vb and F) was carried out. Results: Tumors showed heterogeneously enhanced perfusion, pyruvate, lactate and FDG uptake (Figure 1) with different spatial patterns. Vb and F correlated significantly with total 13C, pyruvate and lactate (p \ 0.05), but not with lactate to total 13C ratio (Figure 2). PET correlated significantly with Vb and F but not with 13C metabolites.

Discussion/Conclusion: The distribution of pyruvate and lactate metabolites, but not the lactate to total carbon ratio, from hyperpolarized 13C-MRSI were found to correlate with tissue perfusion in canine thyroid carcinoma patients. Hence, 13C metabolites can be partially determined by perfusion. This is important for correct physiological interpretation of hyperpolarized 13C-MRSI. Remarkably, the FDG uptake in the tumor did not correlate with 13C metabolites [3]. The relation between hyperpolarized 13C-MRSI and FDG-PET in cancer deserves further investigation. The present study was carried in only three canine thyroid cancer patients. Larger studies and investigations of different cancer types will have to be performed before results can be generalized. References: [1] Gutte, Henrik, et al. ‘‘Simultaneous hyperpolarized 13C-pyruvate MRI and 18F-FDG PET in cancer (hyperPET): feasibility of a new

123

S580 imaging concept using a clinical PET/MRI scanner.’’ American journal of nuclear medicine and molecular imaging 5.1 (2015): 38. [2] Larsson, Henrik BW, et al.’’Measurement of brain perfusion, blood volume, and blood-brain barrier permeability, using dynamic contrast-enhanced T1-weighted MRI at 3 T. ‘‘Magnetic resonance in medicine 62.5 (2009): 1270–1281. [3] Gutte Henrik et al., ‘‘Simultaneous Hyperpolarized 13C-Pyruvate MRI and 18F-FDG PET (HyperPET) in 10 Dogs with Cancer.’’,J Nucl Med 56(11):1786–92 (2015).

618 Metabolic alterations in a depression-like rat model of chronic forced swimming stress: In vivo proton magnetic resonance spectroscopy study at 7T C.-H. Yoo1, K.-H. Song1, S.-I. Lim1, H.-J. Kim1, D.-C. Woo2, B.-Y. Choe1 1 Department of Biomedical Engineering, The Catholic University of Korea, Seoul/KOREA, REPUBLIC OF, 2Asan Institute for Life Sciences, Asan Medical Center, Seoul/KOREA, REPUBLIC OF Purpose/Introduction: The chronic forced swimming stress (CFSS) depression-like animals have been used to investigate the pathophysiology of depression. An in vivo proton magnetic resonance spectroscopy (1H MRS) has been used to investigate metabolic alterations in patients with major depressive disorder (MDD). However, recent in vivo 1H MRS studies were conducted at clinical field strengths which had limitations in the quantification of metabolites with the spectral overlaps. High-field and relatively short-TE MRS can improve the value of MRS in investigating MDD. Thus, the goal of this study was to investigate the CFSS-induced effects in the prefrontal cortex (PFC) of depression-like animals using in vivo 1H MRS at 7T. Subjects and Methods: We used MRS and forced swim tests (FST) to investigate metabolic alterations in the PFC of the rats (n = 13), and the behavioral despairs, by comparing before and after the last CFSS. For the CFSS and FST, the rats were placed in water tank (15/ 5 min; 25/30 C) with the video recording. T2-weighted images were acquired in rat brain to localize the MRS voxel in the PFC region. In vivo 1H MRS was acquired using point-resolved spectroscopy (TR/ TE: 5000/16.3 ms, average: 256, spectral bandwidth: 5000 Hz). Results: CFSS significantly changed swimming behaviors of the rats in FST. Figure 1 illustrates (a-b) axial/coronal MRI scans of the rat brain with (c) the MRS voxel. Figure 2 illustrates (a) in vivo 1H MRS with the results of (b) simulation and (c) in vitro measurements. Although the Glu and Gln spectra obtained with in vivo 1H MRS had slightly lower signal-to-noise ratios, the spectral shapes showed good agreements. The levels of mIns, tCho, NAA, mIns/tCr, and tCho/tCr were significantly increased after the CFSS. Table 1 lists the geometrical overlap ratios, SNRs, and CRLB values.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S581 8. Hasler G, van der Veen JW, Tumonis T, Meyers N, Shen J, Drevets WC. Reduced prefrontal glutamate/glutamine and gamma-aminobutyric acid levels in major depression determined using proton magnetic resonance spectroscopy. Arch Gen Psychiatry, 2007; 64: 193–200.

619 SNR optimization of human brain in vivo at 3 Tesla

31

P MR spectra

A. Manzhurtsev1, O. Bozhko2, T. Akhadov3, N. Semenova4 1 504, N.M. Emanuel Institute of Biochemical Physics of the Russian Academy of Sciences, Moscow/RUSSIAN FEDERATION, 2Radiology, Research Institute of Childen Emergency Surgery and Traumatology, Moscow/RUSSIAN FEDERATION, 3Radiology, Clinical and Research Institute of Emergency Children’s Surgery and Trauma, Moscow/RUSSIAN FEDERATION, 4Dynamic of Chemical and Biological Processes, Semenov Institute of Chemical Physics of RAS (ICP RAS), Moscow/RUSSIAN FEDERATION Discussion/Conclusion: In this study, high-field and short-TE MRS were used to assess CFSS-induced metabolic effects in the PFC in depression-like rats. We confirmed that in vivo 1H MRS could reliably assess metabolic changes. Although significant alterations in Glu and Gln levels were not observed, increased prefrontal mIns, tCho, and NAA levels were partially consistent with the results found in patients with MDD, which suggested that the CFSS-induced behavioral despair and metabolic alterations were similar to those found in human patients with depressive disorders. We expect our findings to contribute to the investigation of the alterations in metabolic systems in MDD, in addition to the monoaminergic system, and provide alternative treatment strategies for patients with MDD. References: 1. Rosen Y, Lenkinski RE. Recent advances in magnetic resonance neurospectroscopy. Neurotherapeutics, 2007; 4: 330–45. 2. Lener MS, Iosifescu DV. In pursuit of neuroimaging biomarkers to guide treatment selection in major depressive disorder: a review of the literature. Ann NY Acad Sci, 2015; 1344: 50–65. 3. Yuksel C, Ongur D. Magnetic resonance spectroscopy studies of glutamate-related abnormalities in mood disorders. Biol Psychiat, 2010; 68: 785–94. 4. Kumar A, Thomas A, Lavretsky H, Yue K, Huda A, Curran J, Venkatraman T, Estanol L, Mintz J, Mega M, Toga A. Frontal white matter biochemical abnormalities in late-life major depression detected with proton magnetic resonance spectroscopy. Am J Psychiatry, 2002; 159: 630–6. 5. Farchione TR, Moore GJ, Rosenberg DR. Proton magnetic resonance spectroscopic imaging in pediatric major depression. Biol Psychiat, 2002; 52: 86–92. 6. Capizzano AA, Jorge RE, Acion LC, Robinson RG. In vivo proton magnetic resonance spectroscopy in patients with mood disorders: a technically oriented review. J Magn Reson Imaging, 2007; 26: 1378–89. 7. Yildiz-Yesiloglu A, Ankerst DP. Review of 1H magnetic resonance spectroscopy findings in major depressive disorder: a meta-analysis. Psychiat Res, 2006; 147: 1–25.

Purpose/Introduction: Phosphorus MR spectroscopy is a unique way of in vivo non-invasive energy and lipid metabolism analysis. Its effectiveness has been demonstrated on MR negative pathology investigation, for example, schizophrenia [1]. The purpose of this study is overcoming of problems related with low SNR of 31P spectra with the help of nuclear Overhauser enhancement and TR optimization in the context of time expeditures. Subjects and Methods: Eleven healthy subjects (aged 18–50, 6 m, 5f) participated in this study. Philips Achieva 3.0T and 31P/1H TX/ RX Rapid Biomed head bird-cage coil was used. The spectroscopic volume of interest sized 80 9 60 9 60 mm was located in brain tissue with the help of Survey sagittal and coronal images and T2 W axial images (*2 min). Voxel position is shown on figure 1 with the red frame, blue frame demonstrates approximate chemical shift displacement artifact for bATP peak.

SV ISIS pulse sequence was used for all data aquisitions with TE = 0.1 ms, FA = 35, NSA = 64 and waltz4 broadband decoupling. TR and Waltz16 NOE mix time (where NOE was enabled) were variated. The study consisted of 2 series: in the first part (*10 min), 3 spectra were acquired with TR = 2, 3 and 4 s. In the second part (*36 min), TR = 3 s was fixed, 11 spectra were obtained with mix time values 0, 250, 500, 750, 1000, 1250, 1500, 1750, 2000, 2250, 2500 ms. Data processing was performed in Spectroview. For the first part, the increase in SNR parameter (in %) for each peak was calculated for 3 pairs of TR (2 s vs 3 s, 2 s vs 4 s, 3 s vs 4 s). For the second part, SNR of each peak for various mix times was normalized on corrsponding values for mix time = 0.

123

S582 Results: The increase of SNR for all compared pairs of TR is demonstrated in table 1.

The examples of SNR behavior as a function of mix time are demonstrated on figure 2. For ATP, PC, GPC, GPX, DN the increase stops at mix time *1500–1750 ms, for PCr, Pi, GPE, PE—at *2250–2500 ms. Maximal NOE increase was 20–35% for different metabolites.

Discussion/Conclusion: TR = 3 s is the most effective value in common, since the threshold against TR = 2 s is waded by most of metabolites, and further increase of TR gives SNR only to Pi and GPE. TR = 4 s is useful for pH (Pi-PCr ppm difference) and lipid analysis. TR = 2 s is the most effective for ATP. SNR gain with NOE ends at mix time *2500 ms, that can technically be used with TR = 3 s and higher. References: 1. A. V. Manzhurtsev, N. A. Semenova, M. V. Ublinskii, T. A. Akhadov, S. D. Varfolomeev//Russian Chemical Bulletin, International Edition, 2016, Vol. 65, No. 6, pp. 1630—1636.

620 Development of a Standardization Phantom for Measuring Brain Gamma-aminobutyric acid (GABA) D. Harasym1, N. Simard1, A. Santos Diaz1, A. Nelson2, M. Noseworthy1 1 School of Biomedical Engineering, McMaster University, Hamilton/ CANADA, 2Kinesiology, McMaster University, Hamilton/CANADA Purpose/Introduction: c-aminobutyric acid (GABA) is the most prevalent inhibitory neurotransmitter in the brain. Due to its

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 importance in both normal brain function and in disease, there is considerable interest in reliable non-invasive measurements in vivo. The concentration of GABA is significantly lower than other dominating metabolites, with brain regions below the 3T detection limit [1]. Thus, a standardized quality assurance (QA) measurement protocol is needed. The purpose of this study was to develop a proton magnetic resonance spectroscopy (1H-MRS) phantom for the standardization of GABA with the goal of long term QA measures. Subjects and Methods: A 20 cm diameter spherical container was used as the housing for five 5 cm diameter metabolite-containing spheres (Figure 1). One of the metabolite spheres was used as a reference based on the ‘‘Braino’’ phantom (GE Healthcare, Milwaukee WI). The other four spheres included 2) Braino + 1 mM GABA (average concentration in the human brain [2]); 3) 1 mM GABA in H2O; 4) Braino + 2 mM GABA; and 5) 2 mM GABA in H2O (Figure 2). Single voxel MEGA-PRESS and PRESS spectra were acquired (Figure 3A) from each sphere in a randomized fashion, with 3 repetitions, during nine sessions. Temperature correction was performed prior to acquisition and spectra with linewidth above 5 Hz were rejected. Each spectrum was fitted using Tarquin [3] to provide measures of metabolite concentrations (Figure 3B-D). System receiver gains, transmit gain, and linewidth were reviewed to ensure that neither system prescan values nor poor spectral acquisition contributed to data variability. Fit quality was assessed using the ratio between the fit residual and spectral noise. A 3-way ANOVA was performed using ‘within session’, ‘between session’ and ‘presence of Braino’ as factors.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S583 and B1 differences over time and how these affect GABA consistency in the QA phantom. References: [1] Puts NA, Edden RA. (2012) In vivo magnetic resonance spectroscopy of GABA: a methodological review. Progr Nucl Magn Reson Spect 60:29–41. [2] Keu Y, Cohen B, et al. (2000) Assessment of GABA concentration in human brain using two-dimensional proton magnetic resonance spectroscopy. Psych Res: Neuroimaging Section 1000:169–178. [3] Wilson M, Reynolds G, Kauppinen RA, Arvanitis TN, Peet AC. (2011) A constrained least-squares approach to the automated quantitation of in vivo 1H magnetic resonance spectroscopy data. Magn Reson Med 65: 1–12.

621 Brown Adipose Tissue Evaluation in Supraclavicular Fat Depot in Patients with Diabetes Mellitus Type 2 using MR-spectroscopy D. Ustyuzhanin1, M. Shariya1, E. Koksharova2, A. Mayorov2, M. Shestakova2, S. Ternovoy3 1 Tomography Department, Cardiology Research Center, Moscow/ RUSSIAN FEDERATION, 2Endocrinology Department, Endocrinology Research Center, Moscow/RUSSIAN FEDERATION, 3 Radiology Chair, Sechenov First Moscow State Medical University, Moscow/RUSSIAN FEDERATION

Results: No statistically significant within (p = 0.2985) or between (p = 0.0707) session differences were noted in ratios of [GABA] (spheres 2:4 and spheres 3:5). There were also no significant differences between spheres with and without ‘Braino’ (p = 0.2448). However, an interaction effect between the ratio of ‘No Braino’ spheres and session was present (p = 0.0042). Discussion/Conclusion: Our GABA phantom is stable and has promise as an option for long term QA. However, the interaction effect indicates the fit should be compared between different fitting programs, such as LCModel and jMRUI. It is important to have a QA protocol for the standardization of GABA measurement thus allowing multi-centre and longitudinal studies. Future work will assess both B0

Purpose/Introduction: Evaluation of brown adipose tissue (BAT) and its role in metabolism remains an important topic in recent researches. Whereas white adipose tissue (WAT) has a lipid storage function, BAT in contrast is involved in energy metabolism and thermogenesis. MR spectroscopy can be used for calculation of fat and water content in fat tissue. Whereas WAT predominately contains fat tissue, BAT water concentration is significantly higher. The most prominent BAT depot is localized in the supraclavicular area. The aim of the study was to evaluate the triglyceride content in the supraclavicular fat depot and in the subcutaneous adipose tissue of the neck in patients with Diabetes Mellitus type 2 (DM2) using MRspectroscopy. Subjects and Methods: 25 patients with DM2 (45.9 ± 10.1 years old, body mass index (BMI) of 31.6 ± 5.4 kg/m2) were included in the study. MR spectroscopy was performed on a 3T human whole body scanner with dedicated 16-channel neck phased array coil at a room temperature of approximately 25 Celsius. Single voxel spectroscopy approach and Point RESolved Spectroscopy pulse sequence were used. The triglyceride content in 3 volumes (left and right supraclavicular fat depots and the subcutaneous adipose tissue of the neck) were determined and compared. Results: The triglyceride content in supraclavicular fat depot varied between 79.2% and 97.1% (mean 92.6 ± 4.2%). The triglyceride content in the subcutaneous WAT of the neck was significantly higher (85.3%-99.3%, mean 95.5 ± 2.9%, p = 0.0007). The triglyceride content in supraclavicular fat depot demonstrated significant moderate correlation with BMI (r = 0.64, p = 0.0009). Results of the study showed significant difference between triglyceride content in supraclavicular fat depot and in subcutaneous adipose tissue of the neck. Lower triglyceride content in supraclavicular fat depot can be explained due to higher water content in brown adipose tissue comparing to white adipose tissue in the neck. Significant correlation between triglyceride content in supraclavicular fat depot and BMI means that patients with higher BAT content in

123

S584 supraclavicular depot had lower BMI. It suggests a positive influence of BAT to prevent obesity in patients with DM2. Discussion/Conclusion: MR spectroscopy can be used for BAT evaluation in patients with DM2. The study results suggest that reducing of BAT content in supraclavicular fat depot can influence on the insulin resistance development in such patients. References: 1. Hamilton G, Smith DL, Jr., Bydder M, Nayak KS, Hu HH. MR properties of brown and white adipose tissues. J Magn Reson Imaging. 2011;34(2):468–73. doi: 10.1002/jmri.22623. 2. Hu HH, Kan HE. Quantitative proton MR techniques for measuring fat. NMR Biomed. 2013;26(12):1609–29. doi: 10.1002/nbm.3025. 3. Cypess AM, Haft CR, Laughlin MR, Hu HH. Brown fat in humans: consensus points and experimental guidelines. Cell Metab. 2014;20(3):408–15. doi: 10.1016/j.cmet.2014.07.025. 4. Cypess AM, Lehman S, Williams G, Tal I, Rodman D, Goldfine AB, et al. Identification and importance of brown adipose tissue in adult humans. N Engl J Med. 2009;360(15):1509–17. doi: 10.1056/NEJMoa0810780. 5. Hamilton, G., Yokoo, T., Bydder, M., Cruite, I., Schroeder, M. E., Sirlin, C. B. and Middleton, M. S. In vivo characterization of the liver fat 1H MR spectrum. NMR Biomed. 2011, 24: 784–790. doi: 10.1002/nbm.1622. 6. Hu HH, Wu TW, Yin L, Kim MS, Chia JM, Perkins TG, et al. MRI detection of brown adipose tissue with low fat content in newborns with hypothermia. Magn Reson Imaging. 2014;32(2):107–17. doi: 10.1016/j.mri.2013.10.003. 7. Izzi-Engbeaya C, Salem V, Atkar RS, Dhillo WS. Insights into Brown Adipose Tissue Physiology as Revealed by Imaging Studies. Adipocyte. 2015;4(1):1–12. doi:10.4161/21623945.2014.965609.

622 Altered white matter structural brain network in patients with subjective cognitive decline and mild cognitive impairment X. Xu1, J.S. Kwan2, H.K. Mak1, A. Wong3, C.T.V. Mok3, E.S. Hui1 1 Department of Diagnostic Radiology, The University of Hong Kong, Hong Kong/CHINA, 2Department of Medicine, The University of Hong Kong, Hong Kong/CHINA, 3Department of Medicine & Therapeutics, The Chinese University of Hong Kong, Hong Kong/ CHINA Purpose/Introduction: To investigate the structural alterations in white matter brain network of patients with subjective cognitive decline (SCD) vs. mild cognitive impairment (MCI) vs. healthy controls using network analysis. Subjects and Methods: We recruited 33 patients with SCD, 29 with MCI and 48 healthy controls. All subjects underwent DWIs with b-values = 1000, 2000 s/mm2 along 32 directions using a 3T scanner (Achieva, Philips). Structural brain network was constructed using tractography and the associated network properties were characterized by Brain Connectivity Toolbox1. One-way ANOVA with Bonferroni post hoc analysis was performed to investigate the group difference in network measures among three groups. All analyses were controlled for age, sex, imaging markers of small vessel disease and brain volume. Results: Compared to controls, subjects with SCD demonstrated significantly higher characteristic path length (p \ 0.001) and lower global efficiency (p \ 0.001) in global network features. MCI patients on the other hand had higher characteristic path length (p \ 0.001), lower global (p \ 0.001) and local efficiency (p = 0.004) than controls, and higher characteristic path length (p = 0.013) than SCD. For

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 local network, significantly decreased clustering coefficient in insula (p = 0.030), post cingulate gyrus (p = 0.006), hippocampus (p \ 0.001) and thalamus (p \ 0.001), as well as decreased nodal efficiency in post cingulate gyrus (p = 0.001), hippocampus (p = 0.002), amygdala (p = 0.001) and thalamus (p \ 0.001) were also observed in SCD patients. Similarly, patients with MCI demonstrated significantly lower clustering coefficient in insula (p = 0.001), post cingulate gyrus (p = 0.017), hippocampus (p \ 0.001), amygdala (p = 0.012) and thalamus (p = 0.008), as well as lower nodal efficiency in post cingulate gyrus (p = 0.003), hippocampus (p = 0.001), amygdala (p = 0.001) and thalamus (p = 0.004).

Discussion/Conclusion: Decreased global efficiency and characteristic path length were demonstrated in SCD patients, although to a lesser extent than MCI, suggesting a decline in inter-regional integrity1. This early involvement of long-range connections in disrupted network configuration has been reported in AD2,3, and further supports the idea that long association fibers are more prone to AD related pathology4. Furthermore, disrupted regional network were found in specific regions, which were frequently identified as early affected sites in preclinical AD studies5-7. In conclusion, network analysis may be able to detect the earliest changes in the AD continuum, before cognitive changes are clearly apparent in the pre-MCI stage. References: 1. Rubinov M, Sporns O. Complex network measures of brain connectivity: uses and interpretations. Neuroimage. 2010;52(3):1059–1069. doi:10.1016/j.neuroimage.2009.10.003. 2. He Y, Chen Z, Evans A, Evans A. Structural Insights into Aberrant Topological Patterns of Large-Scale Cortical Networks in Alzheimer’s Disease. J Neurosci. 2008;28(18):4756–4766. doi: 10.1523/JNEUROSCI.0141-08.2008. 3. Lo CC-Y, Wang P-NP, Chou K-HK, Wang J, He Y, Lin C-P. Diffusion tensor tractography reveals abnormal topological organization in structural cortical networks in Alzheimer’s disease. J Neurosci. 2010;30(50):16876–16885. doi: 10.1523/JNEUROSCI.4136-10.2010. 4. Delbeuck X, Linden M Van der, Collette F. Alzheimer’ Disease as a Disconnection Syndrome? Neuropsychol Rev. 2003;13(2):79–92. doi:10.1023/A:1023832305702. 5. Haroutunian V, Perl DP, Purohit DP, et al. Regional distribution of neuritic plaques in the nondemented elderly and subjects with very mild Alzheimer disease. Arch Neurol. 1998;55(9):1185–1191. http://www.ncbi.nlm.nih.gov/pubmed/9740112. Accessed May 11, 2017.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S585

6. de Jong LW, van der Hiele K, Veer IM, et al. Strongly reduced volumes of putamen and thalamus in Alzheimer’s disease: an MRI study. Brain. 2008;131(12):3277–3285. doi:10.1093/brain/awn278. 7. Zhou Y, Dougherty JH, Hubner KF, Bai B, Cannon RL, Hutson RK. Abnormal connectivity in the posterior cingulate and hippocampus in early Alzheimer’s disease and mild cognitive impairment. Alzheimer’s Dement. 2008;4(4):265–270. doi: 10.1016/j.jalz.2008.04.006.

623 Quantitative Identification of Magnetic Field Inhomogeneities Over Time Due to Physiological Disturbances in the Spinal Cord for MRS Approaches N. Simard School of Biomedical Engineering, McMaster University, Hamilton/ CANADA

Results: During phantom scanning, DB0 field varied by only 2 whereas during the human scanning, DB0 varied by 25. The physiological data demonstrated a correlation between respiration and field inhomogeneities. B1 mapping did not yield physiological conclusions.

Purpose/Introduction: Spinal cord 1H-MR Spectroscopy (1H-MRS) can provide insight into the biochemistry of spine-specific metabolites. As such, this is a promising method for motor research and possibly clinical studies. However, due to the spine’s anatomical location there is a significant degradation of signal quality due to magnetic field inhomogeneities, rendering most MRS approaches inaccurate. Although there has been measurements of B0 in spinal cord MRS [1], we are not aware of the assessment of temporal changes in B0 and B1 relating to physiological disturbances with MRS accuracy. Thus, our goal was to measure temporal changes in B0 and B1 during the length of a typical MEGA-PRESS scan. Subjects and Methods: Experiments were performed using a 3T GE MR750 scanner (General Electric Healthcare, Milwaukee, WI) along with a home designed/built phantom and a healthy human subject. B0 and B1 field maps were each repeatedly acquired over a 10 min timespan in the same location of a spinal cord phantom (Fig. 1) and in the cervical spinal cord of a healthy volunteer (Fig. 2). Each were acquired as a single mid-sagittal slice through the spinal cord, every 10 s for a total of 60 field maps over the 10 min. Pulsed oximeter (cardiac) and respiratory motion were recorder over the length of the 10 min scan at a sampling rate of 25 Hz. Discussion/Conclusion: There are significant links between physiological data and the magnetic field homogeneity within the spinal cord. MRS approaches should continue to employ gating techniques to reduce physiological effects, as it can reduce the amount of noise in the already low signal. In addition, further exploration into the relationship between the magnetic field changes and respiration are needed to possibly correlate the physiology with a protocol to increase speed and performance with regards to spinal cord imaging. References: [1] Cooke F.J. et al. (2004) MRM 51:1122–1128.

123

S586

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

624 Neurochemical modulations in Systemic Lupus Erythematosus with and without neuropsychiatric symptoms by 1H-MR spectroscopy at 3T S. Cuellar Baena1, J. Nystedt2, A. Jo¨nsen3, P. Nilsson3, J. La¨tt4, P. Sundgren1 1 Institution of Clinical Sciences, Department of Radiology, Lund University, Lund/SWEDEN, 2Department of Neurology, Lund University, Lund/SWEDEN, 3Department of Reumatology, Lund University, Lund/SWEDEN, 4Center for Imaging and Function, Ska˚ne University Hospital, Lund/SWEDEN Purpose/Introduction: Systemic lupus erythematosus (SLE) is a chronic inflammatory, immune-mediated disease with clinically distinctive features including non-psychiatric (e.g. stroke, seizures, headaches) and psychiatric (e.g. psychosis, cognitive dysfunction, depression) symptoms. The aim of this study was to determine those neurochemical differences between patients with neuropsychiatric symptoms (NPSLE) and those without them (non-NPSLE) and compare them to a healthy control group (HC) in the frontal white matter by using 1H-MRS at 3T. Subjects and Methods: Magnetic Resonance Spectroscopy (MRS) data from 74 subjects was divided into 3 groups: healthy controls (HC, n = 19), non-Neuropsychiatric Systemic Lupus Erythematosus (non-NPSLE, n = 22) and Neuropsychiatric Systemic Lupus Erythematosus (NPSLE, n = 33). Data for single-voxel spectroscopy (SVS) was acquired at a 3T Magnetom Prisma scanner using PRESS sequence (TE/TR = 30 ms/2 s), VOI = 8 mL positioned in the white matter. Quantification of 7 metabolites (Cr, PCr, GSH, tCho, tNAA, Ins, Glu) was reported using ratios to total creatine (tCr) for this cohort. Overall data quality was uniform across the dataset and 10 spectra were excluded from the original dataset due to poor quality standards. Data quality criteria were set to constrain the LW [0.1 ppm and the SNR B10. Additionally, individual metabolite values were excluded if CRLB B20%. Statistical analysis was performed in SPSS software using one-way ANOVA test for independent variables and Bonferroni post hoc test was used to find the effect of group for each metabolite (p-values \ 0.05). Results: Phosphocreatine (PCr) was significantly increased in nonNPSLE patients compared to HC (ANOVA p = 0.016, Bonferonni p = 0.013), while Glutamate (Glu) was significantly reduced in NPSLE patients compared to HC (ANOVA p = 0.008, Bonferroni p = 0.006). None of the metabolites was significantly different between non-NPSLE and NPSLE patients.

123

Discussion/Conclusion: The present data adds information to previous findings of abnormal metabolic changes in SLE patients. In particular, Glutamate and phosphocreatine may be additional biomarkers specific to white matter pathology to those already seen such as Choline, Glutamine, and N-acetylaspartate of the SLE pathology in the insula (Cagnoli P, et al., 2013). References: Cagnoli P, et al. 2013 Oct;20(10):1286–96. Schmidt-Wilcke T, et al. Neuroimage Clin. 2014 Jul 10;5:291–7. Welsh RC, et al. J Magn Reson Imaging. 2007 Sep;26(3):541–51.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Rise of the Machines 625 Feature selection improves prediction of overall survival from baseline MRI in patients with glioblastoma multiforme M. Ingrisch1, M.J. Schneider1, A. Mittermeier1, C. Pirkl1, B. ErtlWagner2 1 Josef Lissner Laboratory for Biomedical Imaging, Institute for Clinical Radiology, Ludwig-Maximilians-University Hospital Munich, Munich/GERMANY, 2Institute for Clinical Radiology, Ludwig-Maximilians-University Hospital, Munich/GERMANY Purpose/Introduction: Radiomics analyses extract a large number of quantitative image features from each image dataset. Machine learning is used to predict outcome from this feature space. Feature selection, i.e. the extraction of a subset of outcome-relevant features, can speed up the learning process and improve prediction accuracy by eliminating irrelevant and redundant information from the feature space. Here, we investigate the influence of two popular feature selection approaches on the prediction of overall survival from pretherapeutic MRI in patients with glioblastoma multiforme (GBM). Subjects and Methods: 66 patients with GBM were included in this retrospective study. Contrast-enhancing tumor volume was outlined on pre-therapeutic 3D contrast-enhanced T1-weighted MRI images and 208 quantitative image features were derived from each tumor1. Machine learning was performed in R using the mlr framework2. As baseline, a random survival forest (RSF) was trained using all features and 10-fold cross validation. Feature importance was assessed using i) minimum redundancy maximum relevance3 (mrmr), a measure based on mutual information, and ii) the average minimal depth4 (md) of each variable in a RSF. In a grid search, the n most relevant features were selected, with n ranging from 2 to 100. The reduced data were used to train a final RSF using 10-fold cross validation, which predicted an individual risk for each patient. As a measure of prediction accuracy, the concordance index ci was determined. Association of predicted risk with overall analysis was assessed by Kaplan–Meier analysis. High- and low risk groups were determined by splitting on the median of predicted risk, differences between groups were assessed with a log-rank test. p \ 0.05 was assumed to indicate significant difference. Results: The baseline RSF achieved a ci = 0.60 (standard error of the mean sem = 0.067). For mrmre feature selection, the optimal number of features was n = 33, achieving ci = 0.63(0.05). For minimal depth variable selection, the optimal number of features was n = 12, which yielded ci = 0.74(0.045). No significant differences between groups were observed for the baseline approach. Feature selection resulted in improved and significant separation of high- and low risk groups, with p = 0.026 and p \ 0.0001 for the mrme and md approaches, respectively.

Kaplan–Meier analysis of the three feature selection approaches. Groups are not significantly separated in the baseline approach (left),

S587 whereas mrmre and md feature selection are able to stratify patients from baseline MRI. Discussion/Conclusion: Feature selection drastically improves the prediction of overall survival from quantitative image features. The minimal depth approach outperformed the mrmre approach in terms of prediction accuracy and was able to identify all but one long-term survivors. References: 1. Ingrisch M, Schneider MJ, No¨renberg D, et al. Radiomic Analysis Reveals Prognostic Information in T1-Weighted Baseline Magnetic Resonance Imaging in Patients With Glioblastoma. Invest. Radiol. 2017. 2. Bischl B, Lang M, Kotthoff L, et al. mlr: Machine Learning in R. J. Mach. Learn. Res. 2016;17(170):1–5. 3. De Jay N, Papillon-Cavanagh S, Olsen C, et al. mRMRe: an R package for parallelized mRMR ensemble feature selection. Bioinformatics. 2013;29(18):2365–2368. 4. Ishwaran H, Kogalur UB, Chen X, et al. Random survival forests for high-dimensional data. Stat. Anal. Data Min. 2011;4(1):115–132.

626 Machine Learning for Classification of Mild Cognitive Impairment in Parkinson’s Disease Based on Multiparametric MRI at 3T E. Ozturk Isik1, S. Cengiz1, D.B. Arslan1, A. Kicik2, E. Erdogdu3, Z. Tufekcioglu4, B. Bilgic4, H. Hanagasi4, A.M. Ulug1, T. Demiralp2, H. Gurvit4 1 Institute of Biomedical Engineering, Bogazici University, Istanbul/ TURKEY, 2Hulusi Behcet Life Sciences Research Center, Neuroscience Unit, Istanbul University, Istanbul/TURKEY, 3 Psychology and Cognition Research Institute, Bremen University, Bremen/GERMANY, 4Istanbul Faculty of Medicine, Department of Neurology, Behavioral Neurology and Movement Disorders Unit, Istanbul University, Istanbul/TURKEY Purpose/Introduction: Mild cognitive impairment in Parkinson’s disease (PD-MCI) has been reported to be present in 30% of the nondemented patients [1]. A significant proportion of PD-MCI patients evolve into dementia, and early diagnosis of MCI might aid with better patient stratification and preparation for later life events [2, 3]. Previous studies have shown lower cerebral blood flow (CBF) in precuneus [4, 5] and higher choline over creatine (Cho/Cr) in frontal lobe in PD-MCI [6, 7]. Additionally, lower fractional anisotropy (FA) and higher apparent diffusion coefficient (ADC) [8], and changes in resting state connectivity [9] have been reported. This study aims at classification of PD-MCI based on multiparametric MRI using machine learning at 3T. Subjects and Methods: Thirty-nine PD-MCI, 27 cognitively normal PD (PD-CN) and 18 age-matched healthy controls (HC) were included in this study after written informed consent. Multimodal MR imaging was performed at a Philips 3T clinical scanner using a 32 channel head coil. The imaging protocol included T1w and T2w MRI, resting state fMRI (rs-fMRI), diffusion tensor imaging (DTI), arterial spin labeling MRI (ASL-MRI), and proton MR spectroscopic imaging (1H-MRSI). All imaging data were registered to MNI-152 brain atlas and several brain regions were selected. In total, 131 features including age, education level, CBF, connectivity indices, FA, ADC, and spectral peak parameters in various brain regions were used for classification of PD-MCI. First, a data matrix VNxM = 83 9 131 was formed, and non-negative matrix factorization (NNMF) was used to decompose this data matrix into the product of a basis matrix W = 83 9 8 and a coefficient matrix H = 8 9 131 [10]. The

123

S588 z-scores of columns of the basis matrix were used for classification of PD-MCI using Statistics and Machine Learning Toolbox 10.2 in MATLAB. 30-fold cross validation was employed. Results: Figure 1 shows multiparametric MRI data of a 57-year-old PD-MCI patient. Table 1 shows the classification results after NNMF dimension reduction and z-score calculation. Highest accuracy of 85.7% was obtained in classification of PD-MCI versus HC with decision trees (sensitivity = 92%, specificity = 72%). For the classification of PD-CN and HC, ensemble of bagged trees resulted in the best predictive response (accuracy = 75.6%, sensitivity = 89%, specificity = 56%). For PD-MCI versus PD-CN classification, the decision trees had the highest accuracy of 70.8% (sensitivity = 82%, specificity = 56%).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 4. Melzer, T.R., et al., Arterial spin labelling reveals an abnormal cerebral perfusion pattern in Parkinson’s disease. Brain, 2011. 134(Pt 3): p. 845–55. 5. Le Heron, C.J., et al., Comparing cerebral perfusion in Alzheimer’s disease and Parkinson’s disease dementia: an ASL-MRI study. J Cereb Blood Flow Metab, 2014. 34(6): p. 964–70. 6. Griffith, H.R., C.C. Stewart, and J.A. den Hollander, Proton magnetic resonance spectroscopy in dementias and mild cognitive impairment. Int Rev Neurobiol, 2009. 84: p. 105–31. 7. Pagonabarraga, J., et al., Spectroscopic changes associated with mild cognitive impairment and dementia in Parkinson’s disease. Dement Geriatr Cogn Disord, 2012. 34(5–6): p. 312–8. 8. Melzer, T.R., et al., White matter microstructure deteriorates across cognitive stages in Parkinson disease. Neurology, 2013. 80(20): p. 1841–9. 9. Baggio, H.C., et al., Cognitive impairment and resting-state network connectivity in Parkinson’s disease. Hum Brain Mapp, 2015. 36(1): p. 199–212. 10. Lee, D.D. and H.S. Seung, Learning the parts of objects by nonnegative matrix factorization. Nature, 1999. 401(6755): p. 788–91.

627 Automated Pixel-Wise Brain Tissue Segmentation of Diffusion-Weighted Images via Machine Learning A. Ciritsis, A. Boss, C. Rossi Department of Diagnostic and Interventional Radiology, University Hospital Zu¨rich, Zu¨rich/SWITZERLAND

Discussion/Conclusion: In this study, machine learning methods were successfully employed to classify PD-MCI, PD-CN and healthy controls. Highest accuracy was obtained in separating PD-MCI from HC. Future studies will investigate ways of increasing the specificity of the classification algorithms for better discriminating PD-MCI. References: 1. Litvan, I., et al., MDS Task Force on mild cognitive impairment in Parkinson’s disease: critical review of PD-MCI. Mov Disord, 2011. 26(10): p. 1814–24. 2. Hely, M.A., et al., The Sydney multicenter study of Parkinson’s disease: the inevitability of dementia at 20 years. Mov Disord, 2008. 23(6): p. 837–44. 3. Aarsland, D., et al., Prevalence and characteristics of dementia in Parkinson disease: an 8-year prospective study. Arch Neurol, 2003. 60(3): p. 387–92.

123

Purpose/Introduction: The aim of this study was to demonstrate that techniques of machine learning allow for automatic segmentation of brain tissues from diffusion-weighted imaging (DWI) datasets, which are typically acquired in magnetic-resonance (MR) examinations to provide information on tissue perfusion, diffusion restriction and fiber orientation, and to determine the optimal set of DWI-based features for accurate segmentation. Subjects and Methods: Eight healthy volunteers underwent DWI measurements in a 3T MR imager applying a echo-planar imaging sequence with an extensive set of 15 different b-values in the range between 0 and 1300 mm2/s, and 20 encoding directions. From 6 volunteers, the pixel-wise b-value dependent signal decay the signal intensity dependence on the encoding direction, as well as computed mean diffusivity and fractional anisotropy (FA) were used as features to train a support vector machine (SVM) classifier for the 3 classes gray matter, white matter, and cerebrospinal fluid [1, 2]. The datasets of two volunteers were used as validation data to assess accuracy and potential over-fitting. Results: An excellent segmentation result with an accuracy of 82.1% on the validation dataset and 82.2 on the training dataset excluding relevant model over-fitting. Analysis of importance of the features revealed that 4 features from signal intensities at different b-values (5, 10, 500, 1200 s/mm2 normalized to b = 0 s/mm2) and FA were most relevant for classification (Fig. 1). Applying this reduced set of features, an almost identical accuracy could be reached for the validation (training) dataset with 82.2% (81.4%) (Tab 1/Fig 2.)

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S589 Discussion/Conclusion: Applying an optimized DWI acquisition scheme with only slightly longer measurement time compared to standard DTI protocols in brain examinations allows for accurate classification and segmentation of tissues in the brain. References: 1. Cortes C, Vapnik V (1995) Support-vector networks. Machine Learning 20 (3):273–297. 2. Verma R, Zacharaki EI, Ou Y, Cai H, Chawla S, Lee SK, Melhem ER, Wolf R, Davatzikos C (2008) Multiparametric tissue characterization of brain neoplasms and their recurrence using pattern classification of MR images. Acad Radiol 15 (8):966–977.

628 Impact of stratification on supervised classification of subjects with Autism Spectrum Disorders E. Ferrari1, A. Giuliano2, P. Bosco2, P. Oliva3, M.E. Fantacci4, S. Calderoni5, F. Muratori6, A. Retico2 1 Fisica, Universita` di Pisa, Pisa/ITALY, 2Pisa, National Institute for Nuclear Physics (INFN), Pisa/ITALY, 3Physics Department, University of Sassari and INFN, Sassari/ITALY, 4Physics Department, University of Pisa, Pisa/ITALY, 5Developmental Neuroscience, IRCCS Stella Maris Foundation, PISA/ITALY, 6Developmental Neuroscience, IRCCS Stella Maris, Pisa/ITALY

Accuracies of the training and validation datasets for different feature sets. Accuracies [%] selected all features IVIM DTI features Training 82.24 82.15 74.96 72.79 dataset Validation 82.07 81.42 74.86 72.14 dataset

Data sets

Purpose/Introduction: Autism Spectrum Disorders (ASD) are a heterogeneous condition that affects individuals with various degrees of severity. Brain magnetic resonance imaging (MRI) represents a valuable non-invasive technique to study this condition. In literature, there is a great amount of studies based on supervised classification algorithms in order to distinguish subjects with ASD from controls through an analysis of their structural brain images. However, the results obtained are controversial and usually not reproduced on a sufficiently wide statistical sample [2]. Recently, many papers focused on the importance of the stratification of the ASD population to improve classification performance [3, 4]. Subjects and Methods: In this study, 420 morphological features extracted with Freesurfer 6.0 from 1112 subjects from the public database ABIDE I [5] are analyzed. In first place, an outlier analysis has been conducted to exclude the subjects for which the brain segmentation was unsuccessful. Then, six different supervised machinelearning algorithms have been used on different groups of subjects to

123

S590 study the effects of stratification; this has been repeated using different feature normalization methods. Finally, for each classifier we have computed the area under the receiver operating characteristic (AUC) and analyzed the most relevant features for the decisionmaking process. Results: The great amount of train-test experiments we performed allowed us to obtain the following main results: 1) the feature normalization does not have a significant effect on the AUCs; 2) by contrast, the stratification and the type of classification algorithm used have a strong impact. Specifically, whilst the entire group of ASD subjects (according to DSM-5 classification) differed from controls with a modest AUC of 0.62 ± 0.04, once the subjects are subgrouped either according to the DSM-IV-TR separate subcategories of autism, Asperger and Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS), or according to the severity of specific ASD behavior (i.e. ADOS calibrated severity scores), we obtained a maximum AUC of 0.76 ± 0.08. Discussion/Conclusion: We found out that the stratification of the ASD population in more homogeneous subgroups leads to an improvement of the case vs. control classification performance.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Despite the AUC values we obtained are still far away from allowing us to claim for the discovery of a neuroimaging-based biomarker for ASD, the identification of which brain regions are more responsible for the subgroup separations may provide new clues in the understanding of the neurobiology underling the ASD condition. References: [1] Betancur, Catalina. ‘‘Etiological heterogeneity in autism spectrum disorders: more than 100 genetic and genomic disorders and still counting.’’ Brain research 1380 (2011): 42–77. [2] Haar, Shlomi, et al. ‘‘Anatomical abnormalities in autism?.’’ Cerebral Cortex (2014): bhu242. [3] Katuwal, Gajendra J., et al. ‘‘Divide and conquer: sub-grouping of ASD improves ASD detection based on brain morphometry.’’ PloS one 11.4 (2016): e0153331. [4] Alexander, Andrew L., et al. ‘‘Diffusion tensor imaging of the corpus callosum in Autism.’’ Neuroimage 34.1 (2007): 61–73. [5] http://fcon_1000.projects.nitrc.org/indi/abide/.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S591

Paper Posters

Abdominal Imaging - Clinical 629 The dynamics of acute changes in hepatic fat content after dietary manipulation 1

1

1

2

1

M. Hajek , M. Dezortova , P. Sedivy , T. Blahova , M. Drobny , K. Zemankova2, J. Kovar2 1 MR-Unit, Dept. Diagnostic and Interventional Radiology, Institute for Clinical and Experimental Medicine, Prague/CZECH REPUBLIC, 2Experimental Medicine Centre, Institute for Clinical and Experimental Medicine, Prague/CZECH REPUBLIC Purpose/Introduction: The hepatic fat content (HFC) is supposed to vary significantly during the day. It is estimated that 10–20% of dietary fat is transported to the liver (as triglycerides (TG) in remnants or as ‘‘spillover’’ fatty acids). It can be then hypothesized that the administration of high fat load results within a few hours in an increase of liver fat that should be detectable using magnetic resonance (MR) methods. Importantly, the magnitude of such increase is likely to be affected by concurrent changes in glycemia and nonesterified fatty acid (NEFA) concentration. The aim of the study was to determine whether administration of high fat load can induce changes in HFC detectable by MRS methods. Subjects and Methods: Seven healthy insulin sensitive males were included into the study: age 40.7 ± 10.5 years; BMI 26.5 ± 2.9 kg/ m2; glucose 5.4 ± 0.3 mmol/l; insulin 6.1 ± 2.5 mIU/l; HOMA-IR 1.5 ± 0.7; TG 1.7 ± 1.3 mmol/l; cholesterol 4.5 ± 0.3 mmol/l; HFC 2.35 ± 1.52%. Single voxel spectroscopy by PRESS sequence with TR/TE = 4500/30 ms and voxel size VOI = 40 9 30 9 25 mm was used (1 acquisition and three repetition in breath hold). Localization of the voxel was in the 6th liver segment and carefully positioned during subsequent measurement. All spectroscopic data were measured three times and from this we determined reproducibility as 90% (1-SD/Mean). All volunteers underwent 4 experiments differing in the composition of the test meal: Fasting Fat Fat +Glucose Glucose

--150 g of fat (cream) at 0 h 150 g of fat (cream) at 0 h, 3 x 50 g of glucose at 0, 2, and 4 h 3 x 50 g of glucose at 0, 2, and 4 h

Concentration of glucose, triglyceride, insulin, and NEFA was determined 10 times throughout the experiments. The study was performed in accordance with the local ethical committee. Results: The consumption of 150 g of fat (cream) resulted in accumulation of fat in the liver (see Figure). The other dietary interventions did not have any statistically significant effect on hepatic fat accumulation. Individual changes in area under increment curve (AUIC) of hepatic fat were positively correlated with AUIC of NEFA in plasma (see Figure). That supports the idea that NEFA are the major source of liver TG.

Discussion/Conclusion: The results of our pilot study suggest the changes of HFC can be monitored throughout the day by using 1H MR spectroscopy and that changes in HFC correspond to changes of NEFA which represent a principal source for hepatic fat accumulation. References: Supported by Ministry of Health of the Czech Republic: AZV 1628427A and institutional support 00023001 IKEM.

630 Assessment of post radiofrequency ablation of HCC; could MRI subtraction solve M. Rezk1, A. Nazeer2, M. Sonbol2, W. Ali2, A. Osman2, M. Samy1, A. Sabry2, H. Nafady2 1 radiology department, NCI, cairo/EGYPT, 2radiology department, alazhar university, cairo/EGYPT Purpose/Introduction: The aim of this study was to evaluate the role of Dynamic subtraction MRI in the assessement of recurrent or residual tumor viability after radiofrequency ablation. Subjects and Methods: This study was performed in 31 cases of HCC who underwent radiofrequency ablation since May 2015 to October 2016; patients’ ages ranged between 44 and 71 years (mean age 53.1) The study was conducted in our Institute; after approval of the ethical committee; with written consents taken from all patients. All cases had been subjected to full clinical assessment, laboratory investigations with the revision of their previous radiological investigations. Patients were scheduled to undergo MRI within 1 month. Axial T1, T2, T2 SPAIR sequences and dynamic imaging using 3D fat-suppressed T1-weighted gradient echo sequence (THRIVE) were done using 1.5 T Philips machine. The dynamic study was performed after contrast injection consisting of one pre-contrast series followed by four successive post-contrast series. Subtraction imaging was performed by subtracting the pre-contrast series from the late arterial phase. Images procession was performed in the workstation. The images were interpreted by two readers who are experienced in hepatic imaging; the first reader interpreted the dynamic MRI and was blinded to the subtraction MRI images; the second reader interpreted both dynamic and subtraction MRI images. We categorize the patients into two groups: resolved (well ablated) & unresolved residual) groups. the resultant data were statistically analyzed.

123

S592

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

631 Non-Linear Mixed-Effects Modelling Can Reduce the Acquisition Time When Measuring Liver Function Using Gadoxetate-Enhanced MRI M. Karlsson1, M.F. Forsgren1, G. Cedersund2, P. Lundberg1 1 Department of Medical and Health Sciences, Division of Radiological Sciences, Linko¨ping University, Linko¨ping/SWEDEN, 2 Department of Biomedical Engineering, Linko¨ping University, Linko¨ping/SWEDEN

Results: First reader results show 20 lesions (64.5%) were resolved (well ablated) lesions while 11 lesions (35.5%) showed recurrent/ residual viability, Second reader results show 22 lesions (71%) were resolved (well-ablated) lesions while 8 lesions (25.8%) showed recurrent/residual viability. 1 lesion (3.2%) was misregistered. In nonenhanced T1 ablated zone; 4 lesions (12.9%) showed heterogeneous signal and only 2 lesions (6.5%) showed isointense signal while the majority of the ablated lesions about, 25 lesions (80.6%) showed high signal intensity, in T2 ablated zone; 31 lesions imaged after ablation, 2 lesions (6.4%) show heterogeneous low signal, 10 lesions (32.3%) show homogenous low signal and 19 lesions (61.3%) show intermediate signal. No significant correlation between the T2 signal intensity and neoplastic viability was detected. A significant additive value of subtraction to the dynamic study was detected, Discussion/Conclusion: The addition of subtraction technique to dynamic MRI could be a powerful tool in detection of tumor viability especially in lesions having high T1 signal before administration of contrast medium and in turn the appropriate clinical management. References: Abdullah SS, Pialat JB, Wiart M, Duboeuf F, Mabrut JY, Bancel B, Ducerf C, et al. Characterization of hepatocellular carcinoma and colorectal ivermetastasis by means of perfusion MRI. J Magn Reson Imaging 2008; 28:390–5. Akahane M, Koga H, Kato N, Yamada H, Uozumi K, Tateishi R, et al. Complications of Percutaneous Radiofrequency Ablation for hepatocellular Carcinoma: Imaging Spectrum and Management. RadioGraphics 2005; 25:S57–S68. Amit Newatia, Gaurav Khatri and John Hines, et al. Subtraction Imaging: Applications for Nonvascular Abdominal MRI AJR 2007; April Volume 188 number 4, 10.2214AJR.05.2182. Buch SC, Kondragunta V, Branch RA, Brian I, et al. Gender-based outcomes differences in unresectable hepatocellular carcinoma. Hepatol Int. 2008 March; 2(1): 95–101. Catalano A, Anandkumar H, Raul N, Peter F, Cristina R, Dushyant V, et al. Vascular and Biliary Variants in the Liver: Implications for Liver Surgery. RadioGraphics 2008; 28:359–378. Chen G, Ma DQ, He W, Zhang BF, Zhao LQ, et al. Computed tomography perfusion in evaluating the therapeutic effect of transarterial chemoembolization for hepatocellular carcinoma. World J Gastroenterol 2008; 14: 5738–43.

123

Purpose/Introduction: Pharmacokinetic modelling of the hepatic uptake of Gadoxetate, a hepatocyte specific contrast agent, has been proposed for characterizing liver function1-3. However, all of the different approaches have included imaging protocols with long examination time and/or a high temporal resolution. In clinical practice, the imaging protocol should preferably be as short as possible and use the same imaging sequences as a normal Gadoxetate-enhanced MRI. Such sequences are often acquired using breath hold, limiting the temporal resolution. Non-linear mixed-effects modelling is an approach that is widely used in drug development because of its ability to handle data with limited information. Therefore, the aim of this work was to determine if NLME can be used to improve pharmacokinetic modelling of hepatic uptake of Gadoxetate. Subjects and Methods: Thirty-five patients with suspected chronic liver disease were imaged, using a 2-point-Dixon sequence (1.5 T Philips Achieva, TR = 6.5 ms, TE = 2.3/4.6 ms, FA = 13), prior to and up to 30 min, twice every 5 min, after Gadoxetate injection. Regions of interest were placed in the liver, spleen, aorta, and portal vein. Signal intensities were converted into Gadoxetate concentrations. The Gadoxetate concentration time series were analyzed using three different pharmacokinetic models 1-3 (Figure 1). Five different data sets were created by stepwise truncating the data (every 5 min), down to 10 min. The models were parameterized for each data set using a least square fitting. Then the models were parameterized using the NLME approach4 (Figure 2). The parameters from the truncated data were then used to simulate the full data set.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S593

Table 1: The number of patient data sets that failed a chi-2-test. Model Least-Square NLME Forsgren 11 2 Georgiou 13 1 Ulloa 9 6

Results: The parameters that were determined from the truncated data sets were used to simulate the complete 30 min data set. Figure 3 shows results from each model how well NLME improved the prediction of the complete data set. Simulations using short 10 min data sets were evaluated using a Chi2 test on the remaining 20 min of data that were not used to estimate the parameters. Table 1 shows how many of the 35 data sets that failed for each separate model. As can be seen, very few data sets fail when NLME was used.

Discussion/Conclusion: NLME clearly improved the parameter estimates very much. Improvement was observed regardless of which pharmacokinetic model that was used. This has the potential of reducing examination time when characterizing liver function in clinical routine. References: 1 Forsgren. PloS ONE, 9(4):e95700, 2014. 2 Ulloa. NMR Biomed. 23(10):1258–1270, 2013. 3 Georgiou. Invest Radiol. 52:2016. 4 Karlsson. BMC Sys Biol., 9(1):1–15, 2015. 5 Forsgren. ISMRM 2014.

632 RENAL PERFUSION QUANTIFICATION USING RESPIRATORY GATED PSEUDO-CONTINUOUS ARTERIAL SPIN LABELING A. Garcı´a-Ose´s, V. Aramendı´a-Vidaurreta, D. Cano, G. Bastarrika, M. Ferna´ndez-Seara Radiology, Clı´nica Universidad de Navarra, Pamplona/SPAIN Purpose/Introduction: Arterial Spin Labeling is an increasingly popular technique used to measure perfusion, due to its inherent advantages, such as the use of arterial water as a diffusible tracer, unlike gadolinium-based techniques. One of the major challenges to the application of ASL in the body is organ motion. This study presents a respiratory gated pseudo-continuous ASL (pCASL) sequence, applied on the kidney. The goal of the study was to evaluate renal perfusion measurements on healthy volunteers, using the aforementioned method. Subjects and Methods: Six volunteers (age = 25.67 ± 3.27 years) took part in the study, after signing written informed consent. Renal images were acquired on a 3T Siemens scanner with a phased-array abdominal coil and respiratory bellow. Perfusion was measured using pCASL: post-labeling-delay = 1200 ms, labeling duramT/m, Gmaximum/Gaverage = 8, tion = 608 ms, Gaverage = 1 B1average = 1.8 lT. For the readout, bSSFP was used: TR/TE = 1.2/ 1.5 ms, FOV = 350 mm, FA = 508, acquisition-matrix = 128 9 128, FOV = 350–425 mm2, GRAPPA-2. Presaturation pulses were applied at the beginning of each TR. A baseline image was acquired to obtain renal blood flow (RBF) maps. The labeling plane was situated approximately 10 cm above the kidneys. Images were acquired with respiratory triggering. The trigger threshold was set at 10% of the inspiration phase peak, so that imaging was done during expiration (Fig. 1). As a result, TR of sequence relied upon the subject’s respiration rate and triggering accuracy. 50 pairs were acquired to obtain the perfusion signal. To improve alignment, images were registered with Advanced Normalization Tools (ANTs) (1). Mutual Information was used to discard

123

S594

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

badly registered images. RBF maps were obtained using single compartment model (2). ROIs of both kidneys were used to compute mean cortical RBF. Temporal SNR was obtained averaging cortical signal along the perfusion time-series and dividing the result by the standard deviation.

Results: Good quality perfusion maps of both kidneys were obtained in all volunteers (Fig. 2). Cortical RBF values (Table 1) were in concordance with previous studies (3, 4). tSNR increased with TR, except on one subject, where TR was twice the respiration rate (RR), setting the readout on different respiratory phases, affecting tSNR (Fig. 3).

tSNR RBF(ml/min/100g) RR(ms) TR(ms)

Table 1. Results across all volunteers 2.89±1.42 187.84±49.73 4098.3±1125.3 5583.3±1452.6

Discussion/Conclusion: This study shows that respiratory gated pCASL is a feasible method for measuring RBF. Advantages of it include its inherent higher SNR than PASL, which is of uttermost importance given the low fractional volume of blood. Respiration triggering reduces organ motion significantly, and allows freebreathing, which is preferred on certain subjects. References: 1. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee C. NIH Public Access. 2012;54(3):2033–44. 2. Alsop DC, Detre JA, Golay X, Gu¨nther M, Hendrikse J, Hernandez-Garcia L, et al. Recommended implementation of arterial spinlabeled Perfusion mri for clinical applications: A consensus of the ISMRM Perfusion Study group and the European consortium for ASL in dementia. Magn Reson Med. 2015;73(1):102–16. 3. Tan H, Koktzoglou I, Prasad P V. Renal perfusion imaging with two-dimensional navigator gated arterial spin labeling. Magn Reson Med. 2014;71(2):570–9. 4. Sokolska M, Thomas D, Bainbridge A, Golay X, Taylor S, Punwani S, et al. Renal Pseudo-continuous Arterial Spin Labelling (pCASL) MRI: A Repeatability Study. 2014;1–13.

633 MRI-compatible imaging window for longitudinal imaging of mouse ovary F. Bochner1, I. Biton2, M. Neeman1 1 Biological Regulation, Weizmann Institute of Science, Rehovot/ ISRAEL, 2Veterinary Resources, Weizmann Institute of Science, Rehovot/ISRAEL Purpose/Introduction: MRI and microscopy have both been proven to be outstanding tools of preclinical research. MRI enables noninvasive longitudinal imaging of organs and tissues. It provides insight into the tissue structure, metabolism and blood supply both in physiological conditions and pathology. Microscopy performed in live animals enables imaging of cell migration, as well as extracellular matrix remodeling in high resolution. Observing such changes is a key to understand the processes leading to tumor growth and

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 metastasis. Due to development of intravital imaging windows, intravital microscopy can now be also performed repeatedly for prolonged periods of time. Utilization of modern plastic materials enables manufacturing of imaging windows compatible with MRI scanners. Here we present an MRI-compatible imaging window, for longitudinal imaging of mouse ovary. Subjects and Methods: The imaging window was manufactured inhouse according to the design utilized before to make a titanium window. To assure compatibility with MRI we used polyetheretherketone (PEEK). PEEK is a bio-compatible, high-strength thermoplastic material with very good mechanical and chemical properties. The imaging window was surgically implanted into the dorso-lateral side of C57b mouse. One day after, following a short period of recovery, the mouse was imaged in 9.4T scanner (Bruker). T1, T2 and MTR (Magnetization Transfer Ratio) maps were acquired. Corresponding image was acquired with stereomicroscope. It was also possible to utilize the window for longitudinal imaging if ECM remodeling and cell migration after injection of fluorescently labeled ES-2 tumor cells. Results: High resolution image from stereomicroscope provided clues on the location of the organ and its surrounding on T1 and T2 maps. It was possible to correlate the location of the ovary, oviduct and fat pad on both MR and microscopic images. MTR map reflected the distribution of the collagen in the ovary. Images acquired with two-photon microscope demonstrated accumulation and remodeling of collagen correlated with tumor cell proliferation and spread . Discussion/Conclusion: Combination of MRI and intravital microscopy can be very advantageous. Changes in fibrosis content measured with MRI could be correlated in the future with structural properties of collagen imaged with two-photon microscope. Hereby such approach could give an insight into ovarian cancer biology and become a basis for evaluation of novel therapeutic agents. References: Bochner, Filip, et al. ‘‘A novel intravital imaging window for longitudinal microscopy of the mouse ovary.’’ Scientific reports 5 (2015).

634 In vivo magnetic resonance evaluation of potential cancer stem cells isolated from clear cell renal cell carcinoma cell line M. Fiedorowicz1, M.I. Khan2, J. Orzel1, M. Welniak-Kaminska1, D. Strzemecki3, C. Szczylik2, A. Czarnecka2 1 Department of Experimental Pharmacology, Small Animal Magnetic Resonance Imaging Laboratory, Mossakowski Medical Research Centre, PAS, Warsaw/POLAND, 2Department of Oncology with Laboratory of Molecular Oncology, Military Institute of Medicine, Warsaw/POLAND, 3Department of Experimental Pharmacology, Mossakowski Medical Research Centre, PAS, Warsaw/POLAND

S595 Purpose/Introduction: Clear cell renal cell carcinoma (ccRCC) is the most common kidney cancer. Prognosis for ccRCCi s generally poor since it is largely resistant to chemo- and radiotherapy, notably 25–30% of new patients already have metastasis at the time of diagnosis. Numerous data suggest that cancer stem cells/tumor initiating cells (CSCs/TICs) are responsible for development of tumor, disease progression, aggressiveness, metastasis and drug resistance. Their existence has been proven in many tumors including ccRCC [1]. However, tumorigenic potential of CSCs/TICs isolated from established RCC cell lines has never been investigated in vivo. Subjects and Methods: Animal experiments were conducted in accordance to local and EU regulations. CD105+, CD105-, CD44+ and CD44- cells were isolated from Caki-1, RCC cell line. Sorted cells were cultured in FreeStyleTM 293 medium and subsequently injected subcutaneously into NOD SCID mice (5 per group, Charles River, Germany). Additional group of animals received unsorted Caki-1 cells. Mice were scanned with Bruker 7 T MR tomograph. Structural MR images to quantify tumor volume were acquired with T2-weighted TurboRARE and used for both morphological analysis and tumor volume calculation with a region growth algorithm provided by MeVisLab Software (MeVis Medical Solutions AG, Bremen, Germany). Localized MR spectroscopy was performed with PRESS sequence (3 9 3 9 3 mm3 voxels located in the center of the tumors), spectra were analyzed with LCmodel software. Results: All the tested cell subpopulations induced tumor in experimental animals, and all the induced tumors were easily distinguishable in T2-weighted images. However, their growth rate and final tumor volumes were different. Injection of CD44+ cells produced tumors of regular shape that were located under the skin and were similar in terms of morphology, size and growth rate to unsorted Caki-1 cells. Injection of CD105+ produced tumors smaller than unsorted Caki-1. However, CD105+ tumors were penetrating neighboring muscle tissues. Metabolite profiles of the tumors could be divided in two groups: one consisting of unsorted cells, CD105- and CD44+, another one was CD105+ and CD44-. Discussion/Conclusion: We characterized subpopulations of ccRCC cell lines expressing potential biomarkers of CSCs/TICs by in vivo magnetic resonance imaging and spectroscopy. Our results support the hypothesis that CD105 (endoglin) expression might linked to aggressiveness in RCC [2]. References: [1] Cheng, B., et al., Cancer stem cell markers predict a poor prognosis in renal cell carcinoma: a meta-analysis. Oncotarget, 2016. 7(40): p. 65862–65875. [2] Matak, D., et al., Functional significance of CD105-positive cells in papillary renal cell carcinoma. BMC Cancer, 2017. 17(1): p. 21. The study was supported by the Polish National Science Centre (NCN) grant ‘‘OPUS 7’’, grant no. UMO-2014/13/B/NZ1/04010. Project carried out with the use of CePT infrastructure financed by the European Union—the European Regional Development Found in the Operational Programme ,,Innovative Economy’’ for 2007–2013.

123

S596

Cardiovascular, Breast and Chest Imaging 635 Contrast kinetics of Gadobutrol in dynamic contrastenhanced MRI of the breast in patients with histologically proven breast cancers initially classified as BIRADS 5 with and without a computer aided detection system M. Junghans1, D. Roettger2, S.D. Meens-Koreman1, M.E.A.P.M. Adriaensen1 1 Radiology-MRI, Zuyderland MC, Heerlen/NETHERLANDS, 2Image Analysis Ltd, London/UNITED KINGDOM Purpose/Introduction: Compare quantitative enhancement parameters of 1.0 molar Gadobutrol with and without the utilisation of a Computer Aided Detection software in patients with histologically proven breast cancers initially classified as BIRADS-5, undergoing dynamic contrast-enhanced MRI of the breasts. Subjects and Methods: 30 Consecutive patients with histologically proven breast cancers initially classified as BIRADS-5, referred for MRI were scanned with a fixed, complete bilateral breast coverage, protocol including dynamic contrast-enhanced three-dimensional fast low angle shot series with 0.1 ml/kg 1.0 molar contrast on a 1.5 T MRI scanner. After motion correction, maps of the pharmokinetic parameters, i.e. (maximum enhancement (ME), initial rate of enhancement (IRE) and initial rate of washout (IRW)) were calculated. 2 Breast radiologist, blinded for the clinical data, drew a Region Of interest (ROI) in the early subtraction images and in a separate session on the parametric maps images. Results: Of all 31 breast cancer cases, 84% could be analysed after region of interest placement during the two readings by two observers. The Intraclass Correlation Coefficient (ICC) showed an excellent (0.9) agreement in MEmean and IREmean and a fair/good agreement (0.49–0.65) in IRWmean. Maximum diameter was drawn on a fixed early subtraction-, maximum enhancement-, initial rate of enhancement- and area under the curve image showing the breast tumor. The maximum diameter measurement (MDM) using parametric maps correlates excellent (0.9) with the MDM on early subtraction images (p \ 0.001). Discussion/Conclusion: Measurements of contrast kinetics do not significantly differ when the ROI is placed inside a breast tumor with usage of a computer aided detection (CAD) system or when the ROI is placed inside a breast tumor without the use of a CAD-system. The maximum diameter measurement (MDM) using parametric maps correlates excellent (0.9) with the MDM on early subtraction images. Measuring the MDM on parametric maps (ME, IRE, AUC) can reliably be done when using a CAD-system. References: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2441490/Mann, R. M., Kuhl, C. K., Kinkel, K., & Boetes, C. (2008). Breast MRI: guidelines from the European Society of Breast Imaging. European Radiology, 18(7), 1307–1318. Beresford MJ, Padhani AR, Taylor NJ, Ah-See ML, Stirling JJ, Makris A, et al. Inter- and intraobserver variability in the evaluation of dynamic breast cancer MRI. J Magn Reson Imaging. 2006;24(6):1316–25. http://www.ncbi.nlm.nih.gov/pubmed/17058203.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

636 Investigating the relation between circulating cellfree DNA (cf-DNA) level of blood with the DCE-MRI parameters in Breast cancer: a preliminary study P. Sahoo1, T.P. Slavin2, L. Tumyan3, J.N. Weitzel2, L.V. Tongeren4, T. Jensen5, J. Pineda1, R. Rockne1 1 Division of Mathematical Oncology, Beckman Research Institute, City of Hope, Duarte/UNITED STATES OF AMERICA, 2Department of Medical Oncology & Therapeutics Research, Beckman Research Institute, City of Hope, Duarte/UNITED STATES OF AMERICA, 3 Department of Diagnostic Radiology, City of Hope, Duarte/UNITED STATES OF AMERICA, 4Division of Clinical Cancer Genetics, Beckman Research Institute, City of Hope, Duarte/UNITED STATES OF AMERICA, 5Research And Development, Sequenom, Inc, San Diego/UNITED STATES OF AMERICA Purpose/Introduction: Tissue biopsy is the standard diagnostic procedure for confirmation of breast cancers, which is an invasive method. Recent studies have shown that circulating cellfree DNA (cfDNA) can be used as a noninvasive ‘liquid biopsy’ in the management of breast cancer [1]. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is currently the most sensitive modality for the detection of invasive breast cancer [2, 3]. The aim of this study was to compare total circulating cfDNA levels in blood with breast cancer flow kinetics using dynamic contrast enhanced (DCE) MRI in patients with early stage breast cance. Subjects and Methods: This retrospective study examined 14 patients who were diagnosed with breast cancer prior to treatment intervention. Blood samples were collected from each patient, plasma separated using centrifugation, and cfDNA extracted as previously described in [4]. All patients underwent DCE-MRI in a 1.5 T GE scanner. DCE_MRI included 4 dynamic phases (1 pre, 3 post contrast) with temporal resolution of 90 s. Wash-in (slope1) wash-out (slope2) rate were quantified from the dynamic data using in-house developed java based software. High wash-in and wash-out rate regions of interest (ROI) were selected manually by a board-certified radiologist. Mean and standard deviation of wash-in and wash-out rates were computed and used for statistical analysis. Results: The total abundance of cfDNA in peripheral blood (# of copies of cfDNA/mL plasma) showed a positive correlation with wash-in rate and a negative correlation with wash-out rate, however; the correlation was not statistically significant (Figure A, B). A subset of 4 patients with known copy number alterations suggestive of circulating tumor DNA showed a similar trend, however this trend did not reach statistical significance (Figure C, D).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion/Conclusion: A large number of cell types contribute to the total amount of cfDNA found in the peripheral blood. The negative correlation between wash-out rate and total cfDNA could be explained by a number of mechanisms, including increased tumor cell shedding with increased wash-out rate. The result may not have been statistically significant because of the limited sample size and/or the contribution of other factors besides tumor shedding including nontumor DNA to the total amount of cfDNA. The potential correlation of cfDNA with imaging findings could improve detection and early diagnosis of breast cancer, however additional studies are required with larger sample size and additional cfDNA detection methods. References: [1] https://dx.doi.org/10.18632%2Foncotarget.9453. [2] https://doi.org/10.1097/RLU.0000000000001254. [3] https://doi.org/10.1016/j.breast.2016.08.017. [4] https://doi.org/10.1038/npjgenmed.2016.37.

S597 Subjects and Methods: Local Ethical Committees approved the study. Chronic MI (1-month reperfusion) was induced in swine by 90min angioplasty balloon occlusion of the left anterior descending coronary artery. We characterized by contrast-free MRM (T1weighted, T2 map and T2* map sequences, ADC) and correlative histopathology 14 tissue samples from pig infarcted hearts. We built multi-sequence partial least squares discriminant analysis (PLS-DA) models over selected regions of interest (ROIs). Results: Our approach revealed strong statistically significant differences in the internal magnetic properties of different regions in the infarcted heart. The combination of these differences in multi-sequence PLS-DA models allowed a precise classification of ROIs as infarcted, peri-infarcted or remote tissue (specificity 100% and sensitivity 98%). The construction of parametric images permitted the classification of the tissue in a pixel-by-pixel manner.

637 Predictive model of the chronic myocardial infarction based on magnetic resonance microscopy and correlative histopathology I. Pe´rez Terol1, J.M. Morales Tatay2, C. Rios Navarro3, A. Hervas3, A. Ruiz Sauri4, V. Bodi5, D. Monleon6 1 Laboratory of Molecular Imaging and Metabolomic, INCLIVA, Valencia/SPAIN, 2Laboratory of Molecular Imaging and Metabolomic, Universitat de Vale`ncia, Valencia/SPAIN, 3Cardiology, Universitat de Vale`ncia, Valencia/SPAIN, 4Pathology, Universitat de Vale`ncia, Valencia/SPAIN, 5Cardiology, Hospital Clinico Universitario Valencia, INCLIVA, Universitat de Vale`ncia, Valencia/ SPAIN, 6Laboratory of Molecular Imaging and Metabolomic, INCLIVA, Universitat de Valencia, Valencia/SPAIN Purpose/Introduction: Contrast-enhanced cardiac magnetic resonance provide information about myocardial infarct size and microvascular obstruction. These parameters reflect the efficacy of reperfusion therapy and are important for prognosis in survivors of ST-elevation myocardial infarction. Its high tissue contrast make it the method of choice to measure infarct size and determine the transmural extent of infarction. However, the gadolinium contrast show some limitations: little differentiation between the shades of grey that lie in between the black and white contrasted regions and the potential complications because of patient reaction to the contrast agent1. The magnetic resonance microimaging (MRM) is the high resolution version of magnetic resonance imaging (MRI). MRM of excised tissue provides a detailed functional and anatomical picture of identical nature to those obtained by MRI and of higher resolution, which may be used as a bridge between clinical MRI and histopathology made through optical microscopy2. The potential usefulness of MRM and correlative histopathology in human tissue has been demonstrated in brain tumors3. In infarction, this approach may help in detecting new MRI biomarkers of infarction, for more refined delineating of peri-infarcted area and area at risk and for better understanding of the tissue remodeling that take place after infarction, like to remove complications from the use of contrast agents. The aim of this study was to characterize structural and morphological features of the infarcted heart, which affect the magnetic resonance signal in the different imaging sequences.

123

S598 Discussion/Conclusion: Contrast-free multi-sequence MRM provides images with histological detail and accurately delineates the different affected regions in chronic myocardial infarcted. Our study illustrates that MRM and correlative histopathology constitute a potential platform for extracting in the future new cardiac MR biomarkers. References: 1. Beckett KR, et al. Safe Use of Contrast Media: What the Radiologist Needs to Know. RadioGraphics 2015; 35(6):1738–1750. 2. Hervas A, et al. A Multidisciplinary Assessment of Remote Myocardial Fibrosis After Reperfused Myocardial Infarction in Swine and Patients. J Cardiovasc Trans Res 2016; 9(4):321–33. 3. Gonzalez-Segura A, et al. Magnetic resonance microscopy at 14 Tesla and correlative histopathology of human brain tumor tissue. PLoSOne 2011; 6(11):e27442.

638 Single-Breath-Hold Local Pulse-Wave-Velocity Measurement with Quasi Random Sampled k-t-SparseSense MRI

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Imaging-Sequence The MR-sequence was built on a 2D-FLASH sequence provided with additional 1D-motion-encoding gradients (through-plane flow encoding) followed by 10-point DC-readout. The selection of k-space-lines was realized using a low-discrepancy random sequence generating Niederreiter-Numbers and was additionally weighted with a gauss-function to over-represent kspace-center selection [1, 2]. Image Reconstruction and PWV calculation First coil maps were generated using the ESPIRiT-algorithm [3]. The undersampled and reordered PC-k-t-data where subsequently reconstructed iteratevely with k-t-SPARSE-SENSE applying joint sparsity for the different flow encoding data-sets [3, 4]. From the through-plane flow encoded images the cross-sectional area A(t) of the vessel was segmented semi-automatically and the pulse-wave velocity was calculated with the volume flow Q(t) as PWV = dQ/dA [5]. Results: Fig. 1.a shows the SG signal for a typical measurement at the descending aorta. The deviated cardiac gating signal is shown in Fig. 1.b. Based on a sixfold undersampled quasi random acquisition pattern as shown in Fig 1.c. flow encoded images were calculated with k-t-SPARSE-SENSE reconstruction (imaging parameters: Matrix 256 9 174, FOV 300 9 174 mm2, slice-thickness 4 mm, TE 5.6 ms, TR 10 ms). A representative result for the pulse wave velocity calculation from the descending aorta can be seen in Fig 1.d.

V. Herold1, P. Winter2, W.R. Bauer3, P. Jakob1 1 Physics Department, University of Wu¨rzburg, Am Hubland/ GERMANY, 2Physics Department, University of Wu¨rzburg, Wu¨rzburg/GERMANY, 3Medizinische Klinik und Poliklinik 1, Universita¨tsklinikum, Wu¨rzburg/GERMANY Purpose/Introduction: In the present study we provide a new accelerated method to quantify local aortic PWV with self-gated phase-contrast magnetic resonance imaging (PC-MRI). Undersampled data acquisition could be reduced to the duration of one breathhold by using continuously quasi random sampled k-t-Sparse-Sense MRI. Subjects and Methods: Processing of Cardiac SG-Signal. The self-gating (SG) signal was acquired by performing an additional DC (i.e. without phase encoding)—projection along the frequency encoding direction consisting of 10 data points after each k-space line acquisition. Relevant information for cardiac motion was extracted using first principle components of the SG Matrix. Thereupon the individual navigator signals were transformed using a complex wavelet transform at adapted scaling levels for cardiac motion. The resulting phase-signal of the wavelet transform was then applied to directly sort k-space lines into different bins for cardiac motion.

123

Discussion/Conclusion: In this paper we presented a fast MRImethod to quantify local PWV based on highly undersampled selfgated PC-MRI. ECG-signal acquisition can be avoided completely. The high flexibility of data processing would also allow the correction of unstable heart rate during the measurement or further effects of undesirable motion during data acquisition. References: [1] Niederreiter H.; Bull. Amer. Math. Soc. 84 (1978). [2] Herold et al.; Proceedings 24th Annual Meeting ISMRM (2016). [3] Uecker M. et al.; MRM 71:990–1001 (2014). [4] Kim D. et al.; MRM 71:990–1001 (2012). [5] Vullie´moz. et al.; MRM 47:649–654 (2002).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S599

Diffusion Wighted Imaging σ [%]

639 Computer simulations of BSD-DTI versus standard DTI A. Krzyzak, K. Borkowski Geology, Geophysics and Environmental Protection, AGH Univeristy of Science and Technology, Krakow/POLAND Purpose/Introduction: The fidelity of DTI is limited due to a number of factors [1–5]. The important source of errors is the inhomogeneity of the magnetic field gradients [6–8]. The problem can be effectively resolved by applying BSD-DTI technique [9]. On the contrary to standard DTI (S-DTI), it takes into account the spatial variability of the b matrix [6, 10]. In this study, the efficiency of the BSD-DTI and standard DTI was evaluated by computer simulations. Subjects and Methods: The spatial B-matrix (bsd-0) established for field of view FOV-26 9 26 9 26 voxels basing on the diffusion gradient directions. The diffusion gradient distorted by superimposing the spatial pattern function P(r): Gk0 (r) = Gk + GkPk(r) Gl0 (r) = Gl + GlPl(r) Gm0 (r) = Gm + GmPm(r) The pattern functions are: Pk(r) = r(2x2–y2–z2)/475.5 Pk(r) = r(2y2–x2–z2)/475.5 Pk(r) = r(2z2–x2–y2)/475.5 The SD of each pattern is equal to r while its mean value is equal to zero. The B-matrix calculated by equations: bkk = 598.5G2k + 73.5Gk + 12.5 bll = 600G2l + 72Gl + 12.5 bmm = 596G2m + 76Gm + 13. bkl = 597GkGl + 37Gk + 37Gl + 12.5 bkm = 597.5GkGm + 38Gk + 37Gm + 13. blm = 598.5GGm + 37.5Gl + 36Gm + 13. A set of the 6 DTI measurements simulated and the B-matrix discovered in the BSD process (bsd-1). Concurrently, the distribution was indicated with the simplified BSD (bsd-2) [11]. Every DWI’s distorted by Gaussian noise with SD = 0.1%. Moreover, the influence of measurement noise examined by simulating DTI with superimposed noise with SD from 0 to 10%. Results: The eigenvalues and FA obtained for various noise level and uniform B-matrix are shown in table 2. noise [%] 0 2 4 6 8 10

λ1 9.0E-04 8.53E-04 7.56E-04 6.37E-04 5.02E-04 3.60E-04

λ2 1.1E-03 1.13E-03 1.17E-03 1.20E-03 1.23E-03 1.26E-03

λ3 2.0E-03 2.02E-03 2.08E-03 2.16E-03 2.27E-03 2.39E-03

0.00 2.00 4.00 6.00 8.00 10.00

λ1 [mm2/s] 9.00E-04 8.95E-04 8.83E-04 8.66E-04 8.45E-04 8.21E-04

λ2 [mm2/s] 1.10E-03 1.10E-03 1.11E-03 1.11E-03 1.12E-03 1.12E-03

λ3 [mm2/s] 2.00E-03 2.00E-03 2.02E-03 2.03E-03 2.06E-03 2.09E-03

SD(λ1) [mm2/s] 2.36E-06 1.05E-05 2.27E-05 3.71E-05 5.26E-05 6.89E-05

SD(λ2) [mm2/s] 2.34E-06 1.33E-05 2.84E-05 4.47E-05 6.11E-05 7.73E-05

SD(λ3) [mm2/s] 2.36E-06 2.30E-05 4.78E-05 7.52E-05 1.05E-04 1.38E-04

FA

SD(FA)

0.414 0.416 0.421 0.431 0.443 0.457

0.001 0.001 0.001 0.001 0.002 0.002

Discussion/Conclusion: According to table 2 and graphs 1–2 the measurement noise cause systematic error of eigenvalues [2, 3]. As it was shown in Tables 2 and 3 and graph 3 a similar effect is caused by spatial inhomogeneity of the B-matrix. In graph 3 one can notice a shift between eigenvalues of S-DTI, sBSD and BSD-DTI.

FA 0.414 0.429 0.469 0.52 0.584 0.64

Table 3 shows the eigenvalues and FA both with SD for S-DTI, simplified BSD and BSD-DTI. The true FA is 0.414.

123

S600

Acknowledgements: The work was financed by contract No. STRATEGMED2/265761/10/NCBR/2015. References: [1] N. F. Lori, E. Akbudak, J. S. Shimony, T. S. Cull, A. Z. Snyder, R. K. Guillory, and T. E. Conturo, ‘‘Diffusion tensor fiber tracking of human brain connectivity: aquisition methods, reliability analysis and biological results,’’ NMR Biomed, vol. 15, no. 7–8, pp. 494–515, Dec. 2002. [2] A. A. Mariana Lazar, ‘‘Error Analysis of White Matter Tracking Algorithms (Streamlines and Tensorlines) for DT-MRI,’’ 2001. [3] J.-D. Tournier, F. Calamante, M. D. King, D. G. Gadian, and A. Connelly, ‘‘Limitations and requirements of diffusion tensor fiber tracking: an assessment using simulations,’’ Magn Reson Med, vol. 47, no. 4, pp. 701–708, Apr. 2002.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [4] P. Jezzard and R. S. Balaban, ‘‘Correction for geometric distortion in echo planar images from B0 field variations,’’ Magn Reson Med, vol. 34, no. 1, pp. 65–73, Jul.1995. [5] H. Huang, C. Ceritoglu, X. Li, A. Qiu, M. I. Miller, P. C. M. van Zijl, and S. Mori, ‘‘Correction of B0 susceptibility induced distortion in diffusion-weighted images using large-deformation diffeomorphic metric mapping,’’ Magn Reson Imaging, vol. 26, no. 9, pp. 1294–1302, Nov. 2008. [6] A. T. Krzy_zak and Z. Olejniczak, ‘‘Improving the accuracy of PGSE DTI experiments using the spatial distribution of b matrix,’’ Magn Reson Imaging, vol. 33, no. 3, pp. 286–295, Apr. 2015. [7] J. Mattiello, P. J. Basser, and D. Le Bihan, ‘‘The b matrix in diffusion tensor echo-planar imaging,’’ Magn Reson Med, vol. 37, no. 2, pp. 292–300, Feb. 1997. [8] J. Mattiello, P. J. Basser, and D. Lebihan, ‘‘Analytical Expressions for the b Matrix in NMR Diffusion Imaging and Spectroscopy,’’ Journal of Magnetic Resonance, Series A, vol. 108, no. 2, pp. 131–141, Jun. 1994. [9] K. Kłodowski and A. T. Krzy_zak, ‘‘Innovative anisotropic phantoms for calibration of diffusion tensor imaging sequences,’’ Magn Reson Imaging, Dec. 2015. [10] A. T. Krzyzak and K. Klodowski, ‘‘The b matrix calculation using the anisotropic phantoms for DWI and DTI experiments,’’ Conf Proc IEEE Eng Med Biol Soc, vol. 2015, pp. 418–421, Aug. 2015. [11] AT Krzyzak A method for calibrating a diffusion imaging sequence during a DMRI Experiment in an MR Tomograph, PCT/ EP2016/067964, 2016.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Don’t move! Motion Artefacts and Quality Control

S601 Results: For increasing rBw, the artifact size decreases. For increasing TE and susceptibility, the artifact size increases (see figure below).

640 Numerical Simulation of Artifacts in Magnetic Resonance Imaging (MRI) due to Metallic Implants A. Chinnaiyan1, J. Kreutner2, G. Schaefers3 1 Mechanical and medical engineering, Hochschule Furtwangen, Villingen-Schwenningen/GERMANY, 2Research department, MR:comp GmbH, Testing Services for MR Safety and Compatibility, Gelsenkirchen/GERMANY, 3Research department, MRI-STARMagentic Resonance Institute for Safety, Technology and Research GmbH, Gelsenkirchen/GERMANY Purpose/Introduction: In case of MRI examination in close proximity to the implant, it is important to have the test results for the metal artifacts dimensions. Otherwise pathologies might be hidden by the artifact. The MR artifact testing method is addressed by ASTM F2119 [1] which requires experimental MR testing of the artifact for a gradient and spin echo sequence. Although many different parameters and sequences are available the resulting artifact size cannot be easily predicted for each individual sequence. Therefore our goal was to evaluate the precision of available simulation software for artifact determination and support of experimental MR testing. Subjects and Methods: The artifact appearance was evaluated for different parameters to prove the general relation between artifact size and imaging parameters. For simplification, only susceptibility effects have been considered, but there are also other sources of metallic artifacts in MR imaging [2, 3]. Simulation Simulation is done using JEMRIS software [4]. A setup according to the ASTM standard method was created. A titanium rod was defined as a generic test object and placed inside a homogenous phantom filled with medium. A gradient echo sequence was simulated for various parameters (see Table below) like Echo time (TE), receiver bandwidth (rBw) and susceptibility values of the metal, where it is implemented with the help of the multi-component shoulder implant. Parameters TE rBw susceptibility values

Range of values used 5 ms, 15 ms, 25 ms 125 Hz/px, 250 Hz/px, 1000 Hz/px. 3 range of values for each metal alloy component of the shoulder implant

Artifact evaluation The artifact size for each simulated image is calculated according to the ASTM definition [1]. Validation The Validation of the simulation result is done by comparing it with the results of real scanning of the test objects.

123

S602

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 References: [1] ASTM-F2119-07, ‘‘Standard method for the evaluation of MR image artifacts from passive implants’’. 2013. [2] B. Hargreaves, PaulineWorters und K. B. Pauly, ,,Metal induced artifacts in MRI, ‘‘Bd. 197:547–555, Nr. DOI:10.2214/AJR.11.7364, 2011. [3] SM. Aboelmagd, PN. Malcolm und AP. Toms, ,,Magnetic resonance imaging of metal artifact reduction sequences in the assessment of metal-on-metal hip prostheses, ‘‘Bd. 7, Nr. 65–74, 2014. [4] T. Sto¨cker, K. vahedipour, D. Pflugfeld und N. Shah, ,,Highperformance computing MRI simulations, ‘‘Magnetic Resonance in Medicine, Bd. 64(1), pp. 186–193, 2010. [5] John. F. Schenck, ‘‘The role of magnetic susceptibility in MR imaging: MRI magnetic compatibility of first and second kinds,’’ in Medical Physics, Vols. 23, No:6, 1996 June, pp. vol 23, No.6. [6] C. Bakker und R. Bhagwandien, ,,Simulation of susceptibility artifacts in 2D and 3D fourier transform spin-echo and gradient-echo magnetic resonance imaging,‘‘Bd. 12, Nr. 5, 1994.

Discussion/Conclusion: The dependency of the artifact size of different parameters is proved by the simulation, where the applied model considers only susceptibility effects, while neglecting RF and dB/dt induced artifacts. This simplified approach is justified in this case by the good agreement between simulation and experiment. However, extension of the model to additional effects is necessary to achieve a full and flexible artifact assessment for multiple imaging parameters.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

fMRI 641 Arduino based Human Interface Device for fMRI studies J. Rydlo1, L. Hejtmanek2, J. Tintera1, I. Ibrahim1, I. Fajnerova2 1 Radiodiagnostic and Interventional Radiology Department, Institute for Clinical and Experimental Medicine, Prague/CZECH REPUBLIC, 2Applied Neurosciences and Brain Imaging, National Institute of Mental Health, Klecany/CZECH REPUBLIC Purpose/Introduction: For fMRI experiments researchers require a wide range of human interface devices (HID). However, such devices are usually not a standard equipment of MR scanners and commercial devices are often expensive. Therefore, to fulfil requirements in research and clinical fMRI examinations we designed and implemented a new HID system. Our HID system includes a joystick (Fig. 1) and a button response box, which are used inside the MR scanner, and a removable fiber optic bundle and a HID module (Fig. 2), which resides outside the MR scanner room. The joystick and buttons connect to the PC via USB through the HID module and respond just like any other PC compatible joystick, gamepad, mouse or keyboard.

S603 trigger pulses. These pulses are sent by MR scanner to the running research paradigm through USB port. They allow researchers to synchronize MR image acquisition and experiment timestamps. Software: The arduino software is coded using publicly available keyboard, mouse and joystick simulating libraries [2]. These allow to interpret the HID device pulses as keystrokes, or joystick/mouse movement. Therefore no extra programming is required to administer experiments that already use PC keyboard or mouse.

Results: The system can be used for a wide range of computer response tasks used in fMRI studies. Discussion/Conclusion: The device was successfully applied in fMRI study aimed at spatial navigation which required active movement control inside MR scanner. The task designed in Unity3D engine recorded participant’s movement (position and rotation) and actions in the virtual city environment. Each trial required navigation between specific city locations (e.g. hospital and supermarket). Combination of two HIDs was used in order to allow full manipulation of directional control: joystick for rotation and two buttons for forward–backward movement. Third button was used to confirm chosen direction during a pointing task assessed in the beginning of each trial. Supported by Ministry of Health of the Czech Republic, grant nr. 15-28998A and GACR P304-P303-16-167-16729S. References: [1] https://www.arduino.cc/en/Main/ArduinoBoardDue. [2] https://www.arduino.cc/en/Reference/HID.

642 High temporal resolution for BOLD fMRI: a comparison with standard acquisition S. Fall1, J.-M. Constans2, O. Baledent1 1 BioFlow Image, University of Picardy, Amiens/FRANCE, 2 Radiology, Amiens University Hospital, Amiens/FRANCE

Subjects and Methods: Hardware: The HID module (Fig. 3) is based on an open-source physical computing platform Arduino Due [1]. It is equipped with connectors for receiving optical or electrical

Purpose/Introduction: Scan durations of most BOLD fMRI are between 5 and 10 mn with temporal resolutions typically around 2500–3000 ms. However, there is increased interest to achieve faster sampling rates in order to increase sensitivity in event-related fMRI (1), to reduce sensitivity to physiological noise in fMRI, and to infer neural difference in functional network organization. The purpose of this study was to compare BOLD sensitivity between standard acquisition with long repetition time (TR) and rapid acquisition with relatively short TR. Subjects and Methods: Functional images were obtained from two right-handed healthy volunteers with a 3 T-MRI scanner using two different temporal resolutions: (i) TR = 180 ms to acquire two contiguous 4-mm thick axial slices, positioned through the secondary somatosensory cortex: TE = 30 ms, flip angle = 90,

123

S604 FOV = 220 9 235 mm2, matrix = 112 9 117, acquisition time = 1.44 mn, 480 volumes. (ii) TR = 3000 ms, to acquired whole brain images: FOV = 224 9 240 mm2, voxel size = 292 mm2, 50 axial slices, slice thickness = 3 mm, acquisition time = 4 mn, 80 volumes and using the identical TE and flip angle. The functional images consisted of periods of ‘‘baseline’’ alternating with periods of ‘‘swallowing’’. The two functional run were compared based on the mean percent signal change, maximum t-value, the mean t-value and the number of voxels with t [ 3.5 for comparable slices. Results: Significant BOLD changes were detected with the fast acquisition, in concordance with the topography of the activation obtained with the conventional TR (figure 1). The comparison of the BOLD sensitivity between the two TRs is summarized in the table. For both subjects, the short-TR demonstrated higher mean percent signal change, higher mean t-value, higher maximun t-value and more voxels above the threshold than the longer TR. Contributions of fast signal fluctuations from respiratory and cardiac fluctuations can be detected in the frequency spectrum of the time series (figure 2).

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: Our results indicate that shorter TR may provide high sensitivity, allowing to obtain more information that would be undetected with standard acquisition. It was suggested that some connectivity ‘’hubs’’ in brain state were only identified in the high-frequency signal components (2). High-temporal resolution of BOLD fMRI may also help to separate frequencies components (such as respiratory and cardiac fluctuations) and to understand pathological brain activity such as epileptic discharges which are known to be fast and transient. This approach may also be relevant in pre-surgical evaluation of patients with tumors, offering an opportunity to gain significant scan time reductions. References: 1. Posse S, Ackley E, Mutihac R, Rick J, Shane M, Murray-Krezan C, Zaitsev M, Speck O. Enhancement of temporal resolution and BOLD sensitivity in real-time fMRI using multi-slab echo-volumar imaging. NeuroImage 2012;61:115–130. 2. Gohel SR, Biswal BB. Functional integration between brain regions at rest occurs in multiple-frequency bands. Brain Connect. 2015;5:23–34.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Image Reconstruction and Processing 643 Image domain quasi-static background suppression in least square reconstruction for magnetic particle imaging M. Straub, V. Schulz Department of Experimental Molecular Imaging (ExMI)/Physics of Molecular Imaging, Medical Faculty RWTH-Aachen University, Aachen/GERMANY

S605 benchmark, the results of the new method are compared to the established Kaczmarz reconstruction [3, 4]. Thereto, the signal to background ratio in the image domain is analyzed for both reconstruction techniques. For both methods, the same background is used for seeding of the minimizer or subtraction during pre-processing, respectively. Results: The results of the simulation study for the established Kaczmarz algorithm are shown in Fig. 2(a). Fig. 2(b) depicts the result for the here proposed reconstruction technique. The signal-tobackground series are shown in Fig. 3. An enhancement of the signalto-background ratio by a factor of about 10 is visible.

Purpose/Introduction: Magnetic particle imaging (MPI) is a novel tomographic imaging modality which visualizes the distribution of magnetic nanoparticles’ (MNP) based tracer material [1]. Typically, before image reconstruction, the measured MNP signal is corrected for a background signal that is linearly interpolated between two background measurements. The actual reconstruction uses the iterative Kaczmarz algorithm that is equivalent to a least squares reconstruction with nonnegative constraint to enforce positive and real concentrations [2, 3]. However, current MPI devices suffer from a slowly changing background that degrades the imaging quality. This is especially challenging for longer acquisition times. However, frequent interruptions to measure the background are unfeasible as they introduce additional dead time and reposition artefacts. Subjects and Methods: To solve this issue, we propose a novel acquisition and reconstruction scheme in which we acquire an MPI image twice using a small and precisely known spatial displacement. The presumption for our new reconstruction method is that the sample images differ by this spatial displacement, while the background will be unchanged. In order to separate the sample image from the background during reconstruction, we expanded the conventional least square algorithm [4] by introducing the background as a free parameter.

This method is tested on a simulated two-dimensional (see Fig. 1) measurement that includes an actually measured MPI background signal. We chose a shift by one pixel in x and y direction. As a

123

S606

Discussion/Conclusion: The presented image quality improvements in terms of signal-to-background ratio as well as contrast and homogeneity are very promising for the new method. However, due to the used small shift between consecutive measurements, the FOV is reduced. Though, the loss of FOV seems to be outweighed by the image quality improvement. References: [1] Gleich, B. & Weizenecker, J. Tomographic imaging using the nonlinear response of magnetic particles., Nature 435(7046), pp. 1214–1217, 2005. [2] Dax, A. On Row Relaxation Methods for Large Constrained Least Squares Problems, SIAM J. Sci. Comput., 14(3), pp. 570–584, 1993. [3] Weizenecker, J., Gleich, B., Rahmer, J., Dahnke, H. & Borgert, J. Three-dimensional real-time in vivo magnetic particle imaging., Physics in Medicine and Biology, 54(5), L1–L10, 2009. [4] Knopp, T., Biederer, S., Sattel, T. & Buzug, T. M. Singular value analysis for Magnetic Particle Imaging, 2008 IEEE Nuclear Science Symposium Conference Record, IEEE, pp. 4525–4529, 2008.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 computationally expensive and leads to longer reconstruction times, when mapping a large number of NC samples to their corresponding Cartesian locations on the grid [4]. This work aims to accelerate SCGROG for radial acquisitions, using massively parallel architecture of Graphics Processing Units (GPUs). The inherent parallelism in SCGROG gridding is exploited by employing look-up-tables (LUT) based optimized kernels of Compute Unified Device Architecture (CUDA), for significant reduction in image reconstruction time. Subjects and Methods: As an initial step, SC-GROG is implemented on CPU using Visual C ++ platform to sequentially perform selfcalibration and gridding. For GPU-based implementation of SCGROG, LUT based parallelized gridding is performed using three optimized CUDA kernels to reduce the overall computation time of SC-GROG as shown in Figure 1: (i) Kernel_weightsets_comb: precalculate all the possible combinations of 2D gridding weights and store them in LUTs. (ii) Kernel_grid_update_LUTs: each thread calculates the distance of each sample to its closest Cartesian grid neighbor, and retrieves the appropriate pre-calculated 2D weight-set from LUTs, to shift the non-Cartesian signal at the Cartesian location. In this process the value of new calculated point and its location (x and y) on the grid, are stored into LUTs. This is done to preserve the values of all the points shifted at the same Cartesian location. (iii) Kernel_avg: If multiple non-Cartesian points are mapped to the same Cartesian grid point, all the results of the GROG shifts are averaged in parallel.

644 Gridding for Non-Cartesian MRI data using Graphics Processing Units O. Inam1, M. Qureshi2, H. Omer3 1 Electrical Engineering, COMSATS Institute of Information Technology, Islamabad/PAKISTAN, 2Electrical Engineering, COMSATS Institute of Informatio Technology, Islamabad/ PAKISTAN, 3Electrical Engineering, COMSATS Institute of Information Technology, Islamabad, Islamabad/PAKISTAN Purpose/Introduction: Self-calibrating GRAPPA Operator Gridding (SC-GROG) [1–3] is a parameter free, parallel MRI-based gridding algorithm used to shift the non-Cartesian (NC) k-space points to the nearest Cartesian grid locations. However, gridding is

123

To validate the proposed method, the Shepp–Logan phantom image was Fourier-transformed to produce Cartesian k-space data, which was then resampled as radial data by sinc interpolation [5]. For in vivo validation, cardiac MRI data were acquired on 3.0 T scanner (Magnetom Skyra, Siemens). Acquisition details and hardware specification are given in Figure 2.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Results: The computation time of CPU-based SC-GROG and GPUbased SC-GROG, using simulated phantom data with different number of projections is given in Figure 3. An overall speedup (including transfer time between CPU and GPU memory) up to 7.39 is achieved with the proposed method. In the case of 32-coil cardiac dataset, a speed-up of 8.09 is achieved with GPU based implementation of SC-GROG (Figure 3).

S607 Cartesian data gridding with GPU-based SC-GROG provides manifold faster reconstruction than the CPU-based SC-GROG. References: [1]. M. A. Griswold, M. Blaimer, F. Breuer, R. M. Heidemann, M. Mueller, and P. M. Jakob, ‘‘Parallel magnetic resonance imaging using the GRAPPA operator formalism,’’ Magnetic resonance in medicine, vol. 54, pp. 1553–1556, 2005. [2]. N. Seiberlich, F. A. Breuer, M. Blaimer, K. Barkauskas, P. M. Jakob, and M. A. Griswold, ‘‘Non-Cartesian data reconstruction using GRAPPA operator gridding (GROG),’’ Magnetic resonance in medicine, vol. 58, pp. 1257–1265, 2007. [3]. N. Seiberlich, F. Breuer, M. Blaimer, P. Jakob, and M. Griswold, ‘‘Self-calibrating GRAPPA operator gridding for radial and spiral trajectories,’’ Magnetic resonance in medicine, vol. 59, pp. 930–935, 2008. [4]. H. Saybasili, J. A. Derbyshire, P. Kellman, M. A. Griswold, C. Ozturk, R. J. Lederman, et al., ‘‘RT-GROG: parallelized self-calibrating GROG for real-time MRI,’’ Magnetic resonance in medicine, vol. 64, pp. 306–312, 2010. [5]. J. O’sullivan, ‘‘A fast sinc function gridding algorithm for Fourier inversion in computer tomography,’’ IEEE Transactions on Medical Imaging, vol. 4, pp. 200–207, 1985.

645 Line Profile Measure as a stopping Criterion in L +S Reconstruction F. Najeeb, I. Aslam, I. Shahzadi, H. Omer Electrical Engineering, COMSATS Institute of Information Technology, Islamabad, Islamabad/PAKISTAN

Discussion/Conclusion: Highly parallelized low latency GPU-based SC-GROG gridding for radial acquisitions is proposed. The Non-

Purpose/Introduction: L + S reconstruction uses iterative soft thresholding (IST) [1] technique to reconstruct the dynamic MRI data. A major limitation of IST in L + S reconstruction is the right choice of the number of iterations for good reconstruction results. Lesser iterations result in aliasing artifacts while greater iterations result in increased computational cost. This paper proposes a correlation measure of the central line profiles of the reconstructed images in the current and previous iterations as a stopping criterion [2] for IST in L + S approach. The results of the proposed method are compared with conventionally used Bregman distance stopping criterion [3]. Visual assessment, Artifact Power (AP) [2] and Root Mean Square Error (RMSE) [1] are used as performance metrics. Subjects and Methods: Figure-1 shows flow chart of the proposed method. The acquired undersampled dynamic (cardiac) data is fed into L + S reconstruction algorithm. L + S model decomposes the acquired underasmpled data into low-rank ‘‘L’’ (Static) and sparse ‘‘S’’ (Dynamic) matrix components which are iteratively reconstructed using IST as a part of L + S reconstruction [1]. A correlation measure of the central line profiles of the reconstructed images in the current and previous iteratons is proposed as a stopping criterion in this paper. This work uses a correlation value of 0.9999 as a stopping criterion for line profiles correlation measure.

123

S608

Results: The proposed method is tested on MRI short-axis cardiac cine dataset acquired in Cartesian fashion with 3 T Skyra Siemens scanner at Case Western Reserve University, USA, using 30 channel cardiac coils with 11 temporal frames with dimensions (256 9 256 9 30 9 11). Figure-2 shows the L + S reconstruction results obtained using the line profile as a stopping criterion (proposed method) and Bregman distance stopping criterion at acceleration factor (AF) 6 and 8. Table-1 provides AF and RMSE values of the reconstructed images. The results show that the proposed method gives better results both visually and in terms of AP and RMSE, with lesser number of iterations as compared to the Bregman distance stopping criterion.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: The correlation measure of the central line profile acts as an effective stopping criterion for L + S approach. References: 1. Otazo R, et al. Magn. Reson. Med 2015. 2015:1125–1136. 2. M. Khan, et al. Appl. Magn. Reson. 2017: 227–240. 3. B. Liu et al. Magn. Reson. Med. 2009:145–152.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S609

646 Low rank regularization accelerated CS recovery for in vivo high resolution R2* mapping A. Ayaz1, A. Fracasso2, S. Dumoulin2, A.M. Kamboh1, M.W.A. Caan3 1 Neuro-Informatics Lab, School of Electrical Engineering and Computer Sciences, Islamabad/PAKISTAN, 2Spinoza Centre for Neuroimaging, KNAW, Amsterdam/NETHERLANDS, 3Radiology, Academic Medical Center, Amsterdam/NETHERLANDS Purpose/Introduction: High field scanners can image the brain structure at high resolution but with relatively increased scanning time. Multi echo FLASH MR images show high correlation along spatial intra-echo images as well as in inter echo images. This both spatial and temporal sparsity can be exploited to accelerate the imaging using Compressed Sensing (CS) sampling and recovery methods. Thus, we aim to achieve the maximum possible acceleration keeping the details of the images intact. Subjects and Methods: 3D multi-echo FLASH data series were acquired for T2* mapping of a single healthy subject on 7 T scanner with FOV 224 9 224 9 126 mm, resolution 0.7 mm3, echo times ranging from 3 to 21 ms, 6 echoes, TR time 23.4 ms, flip angle 12 degrees and second order image-based B0 shimming. Scans were made with: full sampling; 39 and 49 CS undersampling patterns with variable density poisson discs and a fully sampled elliptical center with a radius of 20 voxels; SENSE undersampling 1.7 9 1.7 (39) and 2.1 9 1.8 (49). Scanning times were 22m30 s, 8m60 s (39), and 6m30 s (49). Fully sampled and SENSE accelerated scans were reconstructed online. CS data were reconstructed offline using the Berkeley Advanced Reconstruction toolbox [1], with ESPIRiT auto calibration [2]. Reconstruction was performed with local low rank regularization [3] along phase-encoding and echo time dimensions with a regularization parameter of 0.01. T2* maps were computed for reconstructed multi-echo data. Results: CS accelerated scans showed minimal blurring as for SENSE accelerated scans, and are comparable to a fully sampled scan with respect to image contrast and details (Fig. 2). Also, T2* mapping of CS accelerated scan has quite clear image details and a fine contrast (Fig. 3).

Discussion/Conclusion: Low rank reconstruction of accelerated imaging is feasible. Faster imaging reduces the likelihood of motion artefacts. Current acceleration factors balance SNR, resolution and scanning time at 7 T. References: 1. Lustig, M. et al. Magnetic resonance in medicine 2010; 64: 457–471. 2. Uecker, M. et al. Magnetic resonance in medicine 2014; 71: 990–1001. 3. Zhang, T. et al. Magnetic resonance in medicine 2015; 73: 655–661.

123

S610

647 FPGA based Architecture for GRAPPA reconstruction M.H.N. Mughal1, O. Inam2, M. Qureshi3, F. Ghazali4, H. Omer5 1 Electrical Engineering, COMSATS Institute of Information and Technology, Islamabad/PAKISTAN, 2Electrical Engineering, COMSATS Institute of Information Technology, Islamabad/ PAKISTAN, 3Electrical Engineering, COMSATS Institute of Informatio Technology, Islamabad/PAKISTAN, 4R & D, NESCOM, Islamabad/PAKISTAN, 5Electrical Engineering, COMSATS Institute of Information Technology, Islamabad, Islamabad/PAKISTAN

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Results: The proposed design is tested using Phantom data on Xilinx 14.5. The Xilinx simulator runs on a PC with an Intel i5-3210 2.5 GHz 4-core CPU with an 8-GB memory. Table 1 shows a comparison of the reconstruction time between CPU and the proposed architecture. The results show that the image reconstruction time, using the proposed architecture is up to 30 times faster than CPU. All the basic modules in this architecture have a latency of 6 cycles and consume 101 + ((mn)(n + 1))/8 cycles to factorize A into Q and R components.

Purpose/Introduction: In this paper, we propose a novel hardware realization of GRAPPA on FPGA to speed up the image reconstruction process. For this purpose, a high performance four stage pipelined architecture for QR Decomposition using modified Gram Schmidt (MGS) method is presented to reduce the computational complexity of estimating reconstruction coefficients in GRAPPA. The results show that the proposed architecture significantly reduces the GRAPPA reconstruction time without compromising the quality of the reconstructed images. Subjects and Methods: GRAPPA [1] is a widely used data driven technique for image reconstruction in parallel MRI. GRAPPA seeks least square fits (x = min||Ax-B||2) during calibration process, to estimate the reconstruction coefficients x, where matrices A and B contain the training data set from the Auto Calibration Signals (ACS). In GRAPPA, a large number of ACS lines are required for accurate calibration and estimation of the missing k-space lines. However, the computation time for solving a largely over-determined system may result in longer reconstruction time. In this work, a novel four stage pipelined architecture is proposed for accurate and efficient estimation of GRAPPA weight sets. The proposed architecture implements low latency QR [2] decomposition module, using three memory banks and four stages to solve the over-determined system of linear equations during calibration phase. Memory banks save the results of matrices Q, R and updated A, whereas the four stages, shown in fig. 2, update these memory blocks sequentially. The proposed architecture uses double precision floating-point operations because single precision or any fixed-point operations may not maintain the integrity of data in GRAPPA reconstruction.

Discussion/Conclusion: FPGA-based low latency four stage pipelined architecture is proposed to accelerate the GRAPPA reconstruction. Experimental results demonstrate that the FPGAbased implementation significantly reduces the GRAPPA reconstruction time without compromising the reconstructed image quality. References: [1]. Griswold MA, J. P. (2002). Generalized auto calibrating partially parallel acquisitions (GRAPPA). Magnetic resonance in medicine, 1202–1210. [2]. Boppana, D., Dhanoa, K., & Kempa, J. (2004). FPGA based embedded processing architecture for the QRD-RLS algorithm. 12th Annual IEEE Symposium on Field-Programmable Custom Computing Machines.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S611

648 Evaluation of compressed sensing reconstruction schemes for 3D radial projections with 23Na MRI P. Polak, M. Noseworthy School of Biomedical Engineering, McMaster University, Hamilton/ CANADA Purpose/Introduction: In vivo 23Na-MRI is desirable due to sodium’s essential role human metabolism (1–2), but its acquisition suffers from many inherent technical challenges. Among these are low gyromagnetic ratio, short T1/T2 times, dedicated transmit/receive coils, non-standard pulse sequences, low signal and long acquisitions (3–4). Compressed sensing (CS) techniques (5) can be utilized in order to improve image quality without increasing scan times. We present here quantitative measures of compressed sensing 23Na-MRI reconstructions in saline phantoms. Subjects and Methods: Experiments were conducted using a GE 3T MR750, using a designed/built birdcage head coil (resonant frequency = 33.7 MHz). A saline phantom was constructed using 8 concentrations of NaCl in distilled water (430, 215… 3.36 mM). Acquisition was via a 3D-radial projection sequence at differing acquisition window lengths (4 ms, 12 ms, 16 ms, and 25 ms). The projections were designed to have a resolution of 3 9 3 9 3 mm3. The other sequence parameters were held identical—TR: 120 ms; FOV: 180 mm; readout bandwidth: ±125 kHz; number of spokes: 11,310; acquisition length: 22:42. Reconstruction utilized the Berkeley Advanced Reconstruction Toolbox (6). Tested reconstructions: non-uniform FFT (NUFFT); conjugate gradient l2-norm, regularization parameters of 25, 50, 100; total variation (TV), regularization of 1. NIFTI files were created from the reconstruction sampled to in-plane resolution of 128 9 128. Results:

Discussion/Conclusion: As expected, the l2-norm and TV reconstructions provide an increase in SNR for each level of undersampling. While the l2/100 gives the best SNR, there is a discernible loss in edge contrast at this level (Figure 2e). Of the methods shown here, the l2/25 appears to give the best trade-off in terms of increased SNR over the NUFFT method with acceptable edge contrast. Unsurprisingly, the best reconstructions were from the acquisitions with the most acquired data, with the CS-methods performance decreasing with the quality of the input data. It is worth noting that the 4 ms sequence acquired 6.25 time less data than the 25 ms sequence—were this sixfold saving in data acquisition translated into a reduced acquisition time, scan time would be \4 min.

123

S612 This work demonstrates the performance and feasibility of CS reconstructions for 23Na-MRI for levels of undersampling. CS methods are essential to overcome the inherent technical challenges for 23Na-MRI. References: 1. Rose AM, Valdes R. Understanding the sodium pump and its relevance to disease. Clin Chem 1994;40:1674–1685. 2. Skou JC, Esmann M. The Na, K-ATPase. J Bioenerg Biomembr 1992;24:249–261. doi: 10.1007/BF00768846. 3. Madelin G, Lee JS, Regatte RR, Jerschow A. Sodium MRI: Methods and applications. Prog Nucl Magn Reson Spectrosc 2014;79:14–47. doi: 10.1016/j.pnmrs.2014.02.001. 4. Nagel AM, Laun FB, Weber M-A, Matthies C, Semmler W, Schad LR. Sodium MRI using a density-adapted 3D radial acquisition technique. Magn Reson Med 2009;62:1565–1573. doi: 10.1002/mrm.22157. 5. Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med. 2007 Dec;58(6):1182–95. 6. Martin Uecker, Frank Ong, Jonathan I Tamir, Dara Bahri, Patrick Virtue, Joseph Y Cheng, Tao Zhang, and Michael Lustig, Annual Meeting ISMRM, Toronto 2015, In Proc. Intl. Soc. Mag. Reson. Med. 23:2486.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Subjects and Methods: Spatially-weighted TV and TGV regularizers were implemented using an ADMM solver9. The different weighting settings were tested using two numerical phantoms. A Sheppard– Logan based phantom, with a simulated magnitude proportional to the susceptibility ground-truth was employed. Second, simulated complex signal from an in vivo COSMOS10 susceptibility map and magnitude data. Gaussian noise was added to each channel of the complex signal and results were evaluated for different SNR levels using five quality metrics. Results: The analytic phantom revealed a substantial improvement in quality scores when using continuous weights for both TV and TGV (Fig. 1).

649 Spatially weighted regularization with Magnitude prior for QSM C. Milovic1, B. Bilgic2, B. Zhao2, J. Acosta-Cabronero3, C. Tejos4 1 Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Santiago/CHILE, 2radiology, mgh, Boston/UNITED STATES OF AMERICA, 3Wellcome Trust Centre for Neuroimaging, University College London, London/UNITED KINGDOM, 4Department of Electrical Engineering, Pontificia Universidad Catolica de Chile, Santiago/CHILE Purpose/Introduction: QSM involves solving an ill-posed problem where susceptibility values are estimated from phase offsets in GRE images. Solutions to this problem have been implemented using regularized optimizations which incorporate prior knowledge about the solution (e.g. gradient sparsity). Additionally, it has been proposed to include the magnitude data into the regularization1,2, relying on the fact that structures in susceptibility maps and magnitude data share similarities. In the context of the Total Variation (TV) regularization, such prior information has been incorporated as a spatially-dependent binary weight that selectively prevents/enables regularization, hence protecting edges of the most prominent features. Such binary mask is typically inferred (thresholding) from the norm of the 3D gradient of magnitude data3,4. In this work, we compared TV regularizations using binary and continuous weights, and explore orientation-dependent weights based on the gradient vector field of the magnitude image. Additionally, we extended this framework to allow Total Generalized Variation (TGV) regularization5–9. TGV consists of two terms, that are associated with the gradient (first order) and curvature of the solution (second order). It remains unclear how spatially varying their balance affects susceptibility reconstructions.

123

Additionally, using weights based on the gradient vector field yielded better metric scores than the norm of the 3D gradient. For the TGV regularization, weighting of the first order term proved to be more relevant, suggesting that weighting the second order term could be ignored. The COSMOS-based phantom confirmed these results, but it also revealed that spatially dependent (continues and binary) weights are more important in low SNR scenarios; for high-SNR experiments, results were largely comparable (Figs. 2, 3).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: Using TV/TGV regularizer with continuous weights improved the quality of QSM reconstructions, especially in low SNR areas. Additionally, it was shown that orientation-aware weights (i.e. the gradient vector field) allow better preservation of edges and noise reduction. Acknowledgments: ACT1416 Programa PIA CONICYT, FONDECYT 1161448, Becas Doctorado Nacional F:21150369. References: 1. Liu T, Liu J, De Rochefort L, Spincemaille P, Khalidov I, Ledoux JR, Wang Y. Morphology enabled dipole inversion (MEDI) from a single-angle acquisition: Comparison with COSMOS in human brain imaging. Magn Reson Med. 2011;66:777–783.

S613 2. Liu J, Liu T, De Rochefort L, Ledoux J, Khalidov I, Chen W, Tsiouris AJ, Wisnieff C, Spincemaille P, Prince MR, et al. Morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map. NeuroImage. 2012;59:2560–2568. 3. Liu T, Xu W, Spincemaille P, Avestimehr AS and Wang Y. Accuracy of the morphology enabled dipole inversion (MEDI) algorithm for quantitative susceptibility mapping in MRI. IEEE Trans Med Imaging. 2012 Mar;31(3):816–24. 4. Wang S, Chen W, Wang C, Liu T, Wang Y, Pan C, Mu K, Zhu C, Zhang X and Cheng J, Structure Prior Effects in Bayesian Approaches of Quantitative Susceptibility Mapping, BioMed Research International. 2016;2738231:10 p. 5. Bredies K, Kunisch K, Pock T. Total Generalized Variation. SIAM J. Imaging Sci. 2010;3:492–526. 6. Knoll F, Bredies K, Pock T, Stollberger R. Second order total generalized variation (TGV) for MRI. Magn Reson Med. 2011;65:480–491. 7. Yanez F, Fan A, Bilgic B, Milovic C, Adalsteinsson E, Irarrazaval P. Quantitative Susceptibility Map Reconstruction via a Total Generalized Variation Regularization. 2013 International Workshop on Pattern Recognition in Neuroimaging. 2013:203–206. 8. Langkammer C, Bredies K, Poser B a., Barth M, Reishofer G, Fan AP, Bilgic B, Fazekas F, Mainero C, Ropele S. Fast quantitative susceptibility mapping using 3D EPI and total generalized variation. NeuroImage. 2015;(2014). 9. Bilgic B., Chatnuntawech I., Langkammer C., Setsompop K.; Sparse Methods for Quantitative Susceptibility Mapping; Wavelets and Sparsity XVI, SPIE 2015. 10. Liu T, Spincemaille P, De Rochefort L, Kressler B, Wang Y. Calculation of susceptibility through multiple orientation sampling (COSMOS): A method for conditioning the inverse problem from measured magnetic field map to susceptibility source image in MRI. Magn Reson Med. 2009;61:196–204.

123

S614

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Interventional MRI 650 Thermographic evaluation of different media types for hot spot detection during a simulated switched MRI gradient magnetic field exposure M. Wolff1, S. Scholz1, A. Douiri1, W. Goertz1, G. Schaefers2 1 Testing laboratory, MR: comp GmbH Services for MR Safety and Compatibility, Gelsenkirchen/GERMANY, 2Research department, MRI-STAR-Magentic Resonance Institute for Safety, technology and Research GmbH, Gelsenkirchen/GERMANY Purpose/Introduction: Gradient-induced heating of active implantable medical devices (AIMD) is a well-known safety issue in MRI. Guidelines to determine the potential heating hazard are specified in [1]. This study includes a comparison of different setups and signals of switched gradient magnetic fields and their suitability to identify hotspots. A titanium disc (Ø = 46.8 mm, h = 0.6 mm) was used as a device under test (DUT) representing the main plane cross section of a commercial AIMD casing. Subjects and Methods: Measurement was conducted within the coil of a pulsed magnetic field simulator (PMFS, MRI-Tec, Germany). The DUT was placed inside a phantom. Temperature rise was measured with an infrared camera (FLIR, USA) positioned above the coil and focussed on the DUTs surface (FIG. 7). Temperature rise was measured while exposing the DUT (covered with Anti-Reflection Coating (AR) in air) to a known pulsed magnetic field. In total, two different media types (gel acc. to [2] and air) were compared while applying a pulsed magnetic field as described in Tab. 1. In addition, a frequency sweep was performed for both signal shapes.

Results: The temperature distribution for both media types are shown in FIG. 1. A circular area with the highest temperature was expected (due to the skin effect) and observed in air with a higher resolution as in the measurement within gel. Both temperature plots show similar characteristics, however the SNR is significantly higher for air as medium (Fig. 2). The sweep measurements (Fig. 4) do not show detectable heating effects in the low frequency range (200–1 kHz) but an exponential growth starting from approximately 1.5 kHz in both media types. This observation is in agreement with previous publications [3].

123

Discussion/Conclusion: Detection of hotspots has proven to be more precise with air as a medium due to the heat insulation characteristics of air. Referred to the hotspot survey in gel, it can be concluded that the heat transfer from the DUT into gel leads to a blurring effect, which impedes the identification of hotspots. Even though this approach does not represent in vivo conditions, the specific locations of hotspots obtained by this method apply to the clinical case. The absence of a gelled saline is therefore to be recommended for additional hotspot detection. The frequency response of the DUT seems to contradict the definition of a worst-case gradient signal given in [1] (Frequency: 270 Hz). This relatively low frequency did not show any heating effects. Further researches regarding the frequency response should be launched. References: [1] Technical specification ISO/TS 10974 ‘‘Assessment of the safety of magnetic resonance imaging for patients with an active implantable medical device’’. [2] ASTM F2182—11a ‘‘Standard Test Method for Measurement of Radio Frequency Induced Heating On or Near Passive Implants During Magnetic Resonance Imaging’’. [3] Sa´nchez et al.: ‘‘MRI Gradient Field-Induced Heating and its Frequency Dependency for Different Materials’’, ISMRM Annual Meeting 2014 Milan, Abstract No. 4498.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S615

Miscellaneous 651 Feasibility study of visualizing annual ring structures of dried wood with Ultra-Short Echo Time (UTE) magnetic resonance imaging (MRI) for chronological measurements M. Mori1, S. Kuhara1, K. Kobayashi1, M. Yamada2, A. Senoo3 1 Department of Medical Radiological Technology, Faculty of Health Sciences, Kyorin University, Mitaka-shi, Tokyo/JAPAN, 2 graduate School of Humanities of Department of Phyllophyte, History and Cultural Studies of History, Tokyo Metropolitan University, Tokyo/JAPAN, 3Graduate School Human Health Sciences of Department of Radiological Sciences, Tokyo Metropolitan University, Tokyo/JAPAN Purpose/Introduction: As one of the non-destructive inspection methods for archaeology, a diagnostic MRI can be used to visualize the annual ring structures for waterlogged and polyethylene glycol (PEG) treated wooden boards. However, it is expected to be difficult if it’s applied to dried wood or PEG treated wooden boards with a very low water content. We have performed a feasibility study of visualizing the annual ring structures of dried wood or PEG treated wooden boards with a very low water content using an Ultra-short TE (UTE) MRI sequence for chronological tree-ring dating. Subjects and Methods: Several samples with different water contents were made from the same softwood (Cryptomeria) or hardwood (Fagus) and used for the evaluations. Conventional sequences were used to obtain T1 and T2 weighted images on a diagnostic 3T MRI (Vantage Titan 3T; Toshiba Medical Systems Corporation) with a 16ch flexible coil. An UTE sequence was also used with four different TE settings (2.0, 1.0, 0.5 and 0.096 ms) to compare the visualization of annual ring structures of woods. Other scan parameters were TR = 3.7 ms, FA = 25 degree and matrix = 352 9 352. Tree-ring curves were made from the MRI images and compared to the curves obtained by conventional method, in which an optical scanner was used. We also tried to visualize the tree-ring structures of wood materials (PEG treated) discovered from the mid-Edo Period remains. Results: An UTE MRI sequence could successfully visualize the annual ring structures of dried wood or PEG treated wooden boards with very low water content, even though the conventional T1 W and T2 W sequences could not visualize anything. Concerning the TE of UTE, the wood with the lowest low-water content could be visualized only with the shortest TE setting (0.096 ms). The resulting tree-ring curves showed good correlations (r = 0.85) with the curves obtained by conventional method with optical scanner. Finally, we could successfully visualize the tree-ring structures of wooden clogs (PEG treated) from the mid-Edo Period remains by using UTE, even if the material had a very low water content. Discussion/Conclusion: The results showed that the UTE MRI sequence could visualize the annual ring structures of dried woods or of those containing only a small quantity of water successfully and it is expected that this method will extend the applicable range of a diagnostic MRI for chronological measurements in the archaeology field.

References: [1] V. Bucur, Nondestructive Characterization and Imaging of Wood (Springer, Berlin, 2003), pp. 215–279.

652 Feasibility study of ultra-high-resolution magnetic resonance imaging (uHR-MRI) for non-destructive tree-ring measurement of archaeological wood M. Mori1, S. Kuhara1, H. Shibou1, K. Kobayashi1, M. Yamada2, A. Senoo3 1 Department of Medical Radiological Technology, Faculty of Health Sciences, Kyorin University, Mitaka-shi, Tokyo/JAPAN, 2graduate School of Humanities of Department of Phyllophyte, History and Cultural Studies of History, Tokyo Metropolitan University, Tokyo/ JAPAN, 3Graduate School Human Health Sciences of Department of Radiological Sciences, Tokyo Metropolitan University, Tokyo/JAPAN Purpose/Introduction: Recently, micro-focus X-CT has used for imaging the tree-rings of ligneous cultural properties and/or the internal structures of carved wood figures to estimate the age of objects. However, it is difficult to apply it for excavated wooden products with high moisture content, because there is not so much difference of CT values between the tree-rings. MRI is expected to have a better discrimination ability of the water contents but the spatial resolution tends to be lower than X-CT. We have performed the feasibility study ofultra-high-resolution magnetic resonance imaging (uHR-MRI), which can obtain extremely high spatial resolution (from 0.02 to 0.05 mm) compared to the conventional MRI, for non-destructive tree-ring measurement of archaeological wood. The accuracy of the obtained tree-ring width curves was also evaluated. Subjects and Methods: The test samples were made from contemporary wood (Chamaecyparis obtuse) and the excavated wooden products of Kamakura era (bottom plate of lunch box, AC 13). Conventional SE sequences were used to obtain T1, T2 and proton density (PD) weighted images (Res: 0.09–0.18 mm) on a diagnostic 3T MRI (Vantage Titan 3T, Toshiba Medical Systems Corporation) with a 16-channnel phased array coil. Ultra-High-Resolution T2 W imaging was performed by greatly reducing the FOV (FOV: 1.2 9 1.2 cm2, TR/TE: 2000/100 ms, matrix: 324 9 384, Res/thick: 0.03 9 0.03 mm2/6 mm, NAQ: 3, acquisition time: 22 min each) and the FOV was repeatedly shifted to cover the whole objects. The treering curves were drawn after stitching all the small FOV images and compared with the curves obtained by the optical scanner. Results: For the T1, T2 and PD weighted images of test samples obtained by conventional sequences, drawing accurate tree-ring curves was difficult. However, for the uHR-T2WI (0.03 mm resolution), highly accurate tree-ring measurement was possible and showed

123

S616 good correlations (r = 0.80) with the tree-ring curves obtained by the optical scanner (2400 dpi). In addition, drawing the tree-ring curves for the excavated wooden products of Kamakura era was also possible by using uHR-MRI. Discussion/Conclusion: The results showed that the proposed method could measure the tree-rings with comparable accuracy as the conventional optical method and/or visualize the internal structure of wood products. This novel method can be applied to the excavated wooden products with high moisture content, instead of X-CT and will further extend the applicable range of historical products for dendrochronology.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Resonance relaxometry and imaging at low and high magnetic fields, respectively, to study the water contents, the phloem and xylem transport in sorghum plants. The combination of these approaches allows us to seek new eco-physiological biomarkers and to design experiments in the laboratory and even in field conditions1. A dedicated NMR device for living plants has been implemented in a climatic chamber which allows a careful control and modification of the environmental parameters during the experimentation over long periods of time2,3. Results: One particular interesting result concerns the investigation of the spatial distribution of water in stems (knot and inter knot) from T1 and T2 MRI 3D images. The modification of the NMR relaxation parameters during dynamic diurnal cycle will be presented in normal and abiotic stress conditions. A direct application could permit to extract eco-physiological biomarkers which allow to explore and model water fluxes during heat and water stresses and analyze their impact on the development of young reproductive organs.

References: [1] Takayuki Okochi, Dendrochronologia 38, pp. 1–10 (2016). [2] V. Bucur, Nondestructive Characterrization and Imageing of Wood (Springer, Berlin, 2003), pp. 215–279.

653 Real-Time dynamical monitoring of plants status in normal and stress conditions: from Low Fields NMR in laboratory to compact NMR in planta R. Sidi-Boulenouar1, C. Coillot1, E. Nativel2, F. Gatineau3, J.L. Verdeil3, C. Goze-Bac1 1 Laboratoire Charles Coulomb Plateforme BioNanoNMRI, Universite´ de Montpellier, Montpellier/FRANCE, 2UMR5214, Institut d’Electronique et des syste`mes (IES), Montpellier/FRANCE, 3UMR AGAP, Centre de Coope´ration Internationale en Recherche Agronomique pour le De´veloppement (CIRAD), Montpellier/ FRANCE Purpose/Introduction: Today, understanding how plants respond to water stress is essential to meet the challenge of developing new cultivars and new irrigation strategies, consistent with the maintenance of crop productivity in the context of global change. In this context, the study of plant/water relations is of central interest for modeling plant and organ responses to biotic and abiotic constraints. Paradoxically, there are very few direct and non-invasive methods to quantify and measure the level and the flow of water in plants. Subjects and Methods: For this purpose, we report on the development of an innovative methodology based on Nuclear Magnetic

123

Discussion/Conclusion: Finally, we present some developments achieved for the optimization of homogeneity of static field, RF antenna, and NMR sequences in order to built a complete portable NMR device working at 333 kHz, with the versatility and thermal conditions to maintain the plant intact. Our ultimate goal is to perform proton NMR experiments directly in the fields. References: 1. MRI of plants and foods, Henk Van As and John van Duynhoven, Journal of Magnetic Resonance 229, 25–34., 2013. 2. NMR probeheads for biophysical and biomedical experiments, Mispelter J, Lupu M, Briguet A, Theorical principles and practical guidelines., London: Imperial College Press, 2006. 3. Signal Modelling of MRI Ribbon Solenoid Coil Dedicated to Spinal Cord Injuries Studies, C. Coillot, R. Sidi-Boulenouar, E. Nativel, E. Alibert, M. Zanca, G. Saintmartin, H. Noristani, N. Lonjon, F. Perrin and C. Goze-Bac, J. Sens. Sens. Syst., 2016. ‘‘This work has been carried out thanks to the support of the LabEx NUMEV project (no. ANR-10-LABX-20) funded by the «Investissements d’Avenir» French Government program, managed by the French National Research Agency (ANR), CIRAD, AGROPOLIS and INRA.’’

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S617

654 Towards an automatization of the ASTM-F2119 standard for MRI compatible needle artefact assessment A. Illanes, J.W. Krug, M. Friebe INKA Intelligente Katheter, Otto-von-Guericke University Magdeburg, Magdeburg/GERMANY Purpose/Introduction: Susceptibility artefacts in magnetic resonance imaging (MRI) can result in a severe degradation of the image quality, hampering its use during an MRI guided intervention. The quantification of artefacts is regulated by the ASTM-F2119 standard [1], which defines an artefact size measurement method before using them in the MRI environment [2]. However, it gives only an approximative method for artefact quantification, leading to high intra- and interobserver variability [3]. Few attempts have been attempted to decrease this subjectivity [4, 5]. This work presents preliminary results of a new approach for a standardized automated assessment. Subjects and Methods: Six MR compatible biopsy needles with outer diameters of 1.0 mm-1.3 mm (ITP, Somatex, Germany) were imaged in a 3T-MRI scanner (Magnetom Skyra, Siemens) using the FLASH sequence parameters shown in Table 1 (Fig. 1a).

Results: Fig. 3 displays the obtained automatic artefact measurements for the six tested needles. Table 2 shows the slice selection by the algorithm and by the experts. An important correlation between manual and automatic artefact measurement in terms of slice selection and artefact length can be observed.

Table 1 MRI sequence parameters. Sequence TR[ms] TE[ms] flip_angle[°] slice_thikness[mm] Resolution[mmxmm] 1 5 2 10 1 1

The proposed approach processed each line profile signal belonging to every slice of a given needle image. At each slice the algorithm first segmented the artefact and generated a search window for the extracted line profiles. They were then smoothed and two reference values computed at each extreme of the signal, which were used for thresholding the artefact using the 30% rule proposed by the standard (Fig. 1b). The line profile artefact values for each slice were then smoothed and the maximum extracted (Fig. 2). Finally the artefact size of the needle was computed by taking the maximum artefact value of all the slices as specified in the standard. Automatic results were compared with manual annotations performed by three image analysis experts using the method presented in [3].

Table 2: Manual and automatic slice selection analysis. Participant N1 N2 N3 N4 N5 1 11 10 12 32 31 2 11 11 11 32 31 3 11 10 11 30 32 Automatic 11 10 11 31 31

N6 32 31 33 32

Discussion/Conclusion: This work presents an alternative option for MRI artefact assessment through an automatization of the main guideline used for testing MRI interventional devices in order to reduce subjectivity between observers. Results show a correct correlation between manual and automatic measurements. References: [1] ASTM, Conshohocken, PA. ASTM Standard F2119-07, Standard test method for evaluation of MR image artifacts from passive implants.

123

S618 [2] Schaefers G. Testing MR safety and compatibility: an overview of the methods and current standards. IEEE Eng Med Biol Mag. 2008;27:23–7. [3] A. Illanes, J. Krug, M. Maatoq, M. Friebe. Does the size of susceptibility artefact assessment—using guidelines—in MRI vary among different users?. Magnetic Resonance Materials in Physics, Biology and Medicine, ESMRMB 2016, 33rd Annual Scientific Meeting, Vol. 29, Supplement 1, pag. S182-S183, 2016. [4] Gu¨ttler F, Heinrich A, Teichgra¨ber U. Software development for the determination of susceptibility artefacts in MRI after ASTM F2119. Biomed. Technik. 2012;57:480–480. [5] A. Heinrich, Ulf K. Teichgra¨ber, and F.V. Gu¨ttler. Measurement of susceptibility artifacts with histogram-based reference value on magnetic resonance images according to standard ASTM F2119. Biomedical Engineering/Biomedizinische Technik 60.6 (2015): 541–549.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Musculoskeletal Imaging 655 Whole-Body(WB) MRI vs Skeletal X-ray in patients with plasma cell disorders: influence of WB-MRI on therapeutic strategies A. Cassara`, D. Negroni, A. Trisoglio, E. Soligo, S. Berardo, L. Sukhovei, G. Leale, A. Carriero, A. Stecco AOU Maggiore della Carita`, AOU Maggiore della Carita`, Novara/ ITALY Purpose/Introduction: The aim of this study is to compare WholeBody MRI (WB-MRI) findings with X-ray and to evaluate the additional diagnostic role of DWIBS sequences in patients with plasma cell disorders. An additional purpose is to determine the influence of WB-MRI on therapeutic management. Subjects and Methods: We restrospectively reviewed 130 WB-MRI from January 2013 to January 2017. We considered only patients with a diagnosis of plasma cell disorders (Smoldering Myeloma-SM, Solitary Plasmacytoma-SP and Multiple Myeloma-MM) who underwent X-ray Survey and WB-MRI up to 180 days, without therapy between the two exams and with no other malignancies. 39 patients with a mean age of 66.1 ± 11.7 years (23 males and 16 females) were included: 3 SP, 11 SM, 25 MM (15 new diagnosis, 6 complete remission and 4 partial remission). Two readers, in consensus, reports the number of lesions in each region on X-ray Survey and WB-MRI (morphological and DWIBS sequences). The disease was staged according to Salmon-Durie and Salmon-Durie PLUS. We conduct a ‘‘per site’’, ‘‘per region’’ and ‘‘per patient’’ analysis using Coehn’s K test and Symmetry test by Bowker. Results: 585 sites were compared: 43 were positive on X-ray vs 218 on WB-MRI and 209 on DWIBS sequences. The differences between X-ray and WB-MRI were statistically significant (p \ 0.05) in cranium, ribs, right humerus and right scapula. There was no statistatical differences between morphological and DWIBS sequences. Based on WB-MRI, tumor stage was upgraded in 12 of the 39 patients. In 5 of 12 patients, the WB-MRI results allowed the start of therapy; in 2 of 12 the WB-MRI influenced the choise, delaying the treatment. Discussion/Conclusion: WB-MRI strongly influences therapeutic strategies, also without contrast media injection. This tool is able to detect pre-lytic lesions, before bone parenchyma lose enough to be visble at X-ray. This non-invasive technique should be introduced in the routinary study of patients with plasma cell disorders both as a tool to increase the number of patients need to treat and able to influence the quality of patient’s life. References: Christina Messiou, M. K. (2015). Whole body diffusion weighted MRI—a new view of myeloma. British Journal of Haematology. Fre´de´ric E. Lecouvet, M. P. (2016). Whole-Body MR Imaging: Musculoskeletal Applications. Radiology. Metzgeroth, D. J. (2009). Comparison of whole-body MR imaging and conventional X-ray examination in patients with multiple myeloma and implications for therapy. Ann Hematol. S. Rajkumar, M. D. (2014). International Myeloma Working Group updated criteria for the diagnosis of multiple myeloma. The Lancet Oncology. S. Raza, S. S. (2017). The Critical Role of Imaging in the Management of Multiple Myeloma. Current Gematologic Malignancy Reports.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

656 MR diagnosis of entrapment neuropathies at the wrist S. Mehra, A. Singhal Radiodiagnosis, Pgimer Dr. Ram Manohar Lohia Hospital, New Delhi/INDIA Purpose/Introduction: The present study was conducted in the Department of Radio diagnosis, PGIMER Dr. RML Hospital. The purpose of our study was to characterize the structural changes at the wrist joint in entrapment neuropathy by MRI, to assess the role of MR in diagnosing and differentiating the various etiologies of entrapment neuropathy at the wrist . Subjects and Methods: We evaluated a total of 31 patients who were clinically suspected with or who had electrodiagnositc studies suggestive of entrapment neuropathy at the wrist. The study included 27 females and 4 males. The patients underwent MR examination of the wrist using a multichannel extremity flex coil. Results: MRI was capable of characterizing the structural changes at the wrist joint and could detect the morphological abnormalities within the carpal tunnel as well as Guyon’s canal. MR imaging of wrist was able to diagnose entrapment neuropathy in all 31 patients who were symptomatic by evaluation of characteristic changes in the median nerve and/or ulnar nerve cross sectional area, signal intensity and abnormal shape. The nerve that was seen most commonly affected at the wrist joint was median nerve (involved in all 31 patients). Ulnar nerve entrapment was found in 2 patients in our study. Increased signal intensity of median nerve on T2/PD weighted image was present in 13 patients. 18 patients in our study showed normal signal intensity of the involved nerve with reduced cross sectional area at the level of hamate indicating chronic nerve compression. Abnormal enhancement of the affected nerve was found in all 31 patients. We were able to evaluate the flexor retinaculum accurately in all wrists. Abnormal bowing of flexor retinaculum was found in 38.70% of the patients. Discussion/Conclusion: Compression of the median nerve in the carpal tunnel leads to CTS and of ulnar nerve in the Guyon’s canal produces ulnar tunnel syndrome. Impingement upon the nerve by adjacent anatomical structures within these fibro osseous canals result in chronic mechanical abrasion. MR is an excellent modality for accurate diagnosis of entrapment neuropathies, enabling high quality imaging of the nerves within the carpal tunnel and Guyon’s canal. MR imaging was able to diagnose entrapment neuropathy in all 31 patients by evaluation of characteristic changes in the nerve cross sectional area, signal intensity and abnormal shape. MRI provided additional information about synovial involvement and carpal bone erosion thereby allowing accurate etiological diagnosis. Based on MR findings we could accurately diagnose CTS in all 31 patients and Guyon’s canal syndrome in 2 out of 31 patients. References: Middleton WD, Kneeland JB, Kellman GM, Cates JD, Sanger JR, Jesmanowicz A, et al. MR imaging of the carpal tunnel: normal anatomy and preliminary findings in the carpal tunnel syndrome. AJR Am J Roentgenol. 1987 Feb;148(2):307–16. Andreisek G, Crook DW, Burg D, Marincek B, Weishaupt D. Peripheral neuropathies of the median, radial, and ulnar nerves: MR imaging features. Radiogr Rev Publ Radiol Soc N Am Inc. 2006 Oct;26(5):1267–87. Bordalo-Rodrigues M, Amin P, Rosenberg ZS. MR imaging of common entrapment neuropathies at the wrist. Magn Reson Imaging Clin N Am. 2004 May;12(2):265–279.

S619

Neuroimaging - Clinical 657 WITHDRAWN 658 WITHDRAWN 659 Quantitative MR and diffusion tensor parameters in normal appearing brain of glioma patients are correlated with age F. Raschke1, T. Wesemann2, H. Wahl2, S. Appold3, M. Krause4, J. Linn2, E. Troost4 1 partner site Dresden, National Center for Tumor Diseases, Dresden/ GERMANY, 2University Hospital Carl Gustav Carus and Medical Faculty of Technische Universita¨t, Dresden, Institute of Neuroradiology, Dresden/GERMANY, 3Department of Radiotherapy and Radiooncology, University Hospital Carl Gustav Carus and Medical Faculty of Technische Universita¨t, Dresden, Dresden/ GERMANY, 4OncoRay-National Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, Technische Universita¨t Dresden, Helmholtz-Zentrum DresdenRossendorf, Dresden/GERMANY Purpose/Introduction: Quantitative MRI (qMRI) and diffusion MRI (dMRI) hold potential for assessing normal and abnormal tissue characteristics and early treatment related changes in glioma patients, using quantitative image intensities. Here we assess if variations in qMRI/dMRI measures of normal appearing brain (NAB) of glioma patients are related to patient age and thus need to be taken into account in comparative analyses. Subjects and Methods: Twelve glioma patients (8 males, 32–80 years: two grade II, six grade III, four GBM, t test: no significant age difference between grades) were scanned as part of an ongoing study after resection/biopsy before radio(chemo)therapy. Data was acquired on a 3T Philips scanner using an 8 channel head coil: 3D gradient spoiled echo (FA = 3/20, TR/TE = 10/3.7 ms, 1 9 1 9 1 mm3), B1 mapping [1], 2D-FFE (TR = 1355 ms, slice thickTE = 5.8/9.1/12.4/15.8/19.1 ms, 2 9 2 mm2, ness/gap = 2/1 mm), 2D-TSE (TR = 5.2 s, 10echos, TE1/ DTE = 18.6/30 ms, 0.45 9 0.45 mm2, slice thickness/gap = 4/ 1 mm), DTI (b = 1000 s/mm2, 32 directions, 2 9 2 9 2 mm3, TR/ TE = 6.5 s/66 ms), 3D-FLAIR (0.5 9 0.5 9 0.5 mm3). Calculated qMRI/dMRI parameters (Figure 1): T1/PD [2]; T2/T2* via monoexponential decay; FA, MD, k1 (axial diffusion), kRD (radial diffusion) via FSL. Grey matter (GM) and white matter (WM) maps generated with SPM12 from T1maps, excluding abnormal tissue and the cerebellum, were used to calculate GM and WM histograms and their maxima. Regression analysis was performed in MATLAB using Pearson’s correlation with a significance threshold of p = 0.05.

123

S620

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 In conclusion, despite numerous potential confounding elements, age is a significant factor related to qMRI/dMRI changes in NAB of glioma patients with similar trends compared to normal aging and should be taken into account when performing comparisons. References: [1] Cunningham et al. Magn Reson Med 2006;55:1326–1333. [2] Volz et al. Neuroimage 2012;63:540–552. [4] Cox et al. Nature communications 2016;7:13629. [5] Sedlacik et al. NeuroImage 2014;84:1032–1041. [6] Zhang et al. NeuroImage 2012;61:1000–1016. [7] MacKay et al. Magnetic resonance imaging 2006;24:515–525.

Results: In WM, T1, PD, T2, MD, k1, kRD increase statistically significantly with age, whereas FA decreases (Figure 2). In GM only kRD increases with age, the other parameters show no agedependence.

Discussion/Conclusion: Diffusion [4] and T2 trends [5] in WM are in line with those seen in normal aging. Here, WM changes are likely driven by an increasing water pool with low diffusion restriction as derived from PD, T2, MD, k1, kRD. More sophisticated analysis such as NODDI [6] and T2 NNLS [7] are needed to investigate if FA reductions are solely caused by this increased MD or partly due to demyelination, axonal loss and/or axonal dispersion. We observed no GM T2 changes with age, in contrast to previously reported GM T2 increase in normal aging [5]. CSF related partial volume effects due to atrophy can easily bias mean and median GM qMR and dMR values, which were reduced in our study by using maximum histogram values (see Figure 3).

123

660 Diffusion kurtosis imaging values along interhemispheric and associative fiber tracts in healthy volunteers E. Pogosbekian1, E. Sharova2, N. Zakharova3, L. Fadeeva3, I. Maximov4, I. Pronin3 1 Neuroimaging, Burdenko Neurosurgical institute, Moscow/RUSSIAN FEDERATION, 2Laboratory of General and Clinical Neurophysiology, Institute of Higher Nervous Activity and Neurophysiology RAS, Moscow/RUSSIAN FEDERATION, 3 Neuroimaging, Burdenko Neurosurgical Institute’’, Moscow/ RUSSIAN FEDERATION, 4Experimental Physics III, TU Dortmund, Dortmund/GERMANY Purpose/Introduction: In our previous studies [1] we investigated regional specificity of correlation between corpus callosum (CC) tracts damage and degradation of consciousness state in patients with severe traumatic brain injury (STBI). We evaluated diffusion tensor imaging (DTI) parameters in 7 CC regions. We suppose that anterior commissure (AC) tracts can compensate lack of CC ones in case of STBI [2]. In other studies revealed correlation of consciousness level in patients with STBI and EEG synchronization between the anterior and posterior cortex correlated [3], so we suggested that long associative tracts integrity is strongly related with consciousness state of the patients. Diffusion kurtosis imaging (DKI) is a modern approach, which is sensitive to changes in white matter [4]. Purpose of this study is to achieve DKI and DTI values for healthy volunteers along CC (divided into 7 segments), Inferior fronto-occipital fasciculus (IFOF) and AC tracts. These values can be useful in other studies of fiber tracts integrity in patients with STBI as reference ones. Subjects and Methods: In this study participated 13 healthy volunteers (8 men, 5 women; age from 21 to 55, average 34). We used 3,0 Tesla scanner, voxel size 3 9 3 9 3 mm, FOV 240 mm, b-values were 1000 and 2500 s/mm2, 60 diffusion gradient directions for each b-value. The same volumes were used for HARDI CSD fiber tracking [5]. CC (7 segments), IFOF and AC fiber tracts were constructed by two independent raters for each volunteer. DKI and DTI values were measured along these tracts. Results: are presented in table. Order of CC segments numbering: from posterior to anterior. Inter-rater correlation was 0.96 (p \ 0.001).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

AK AC IFOF CC 1 CC 2 CC 3 CC 4 CC 5 CC 6 CC 7

AWF

AxEAD*10-

FA

3

0.62

0.62

1.71

0.64

0.29

1.75

0.61

0.3

2.21

0.67

0.3

2.16

0.71

0.32

0.69

0.3

0.68

0.21

KA

MD*10-3

MK

S621

RadEAD*10-

RK

3

TORT

0.11

1.12

0.71

1.3

1.3

1.31

0.17

0.99

0.81

1.11

1.02

1.58

0.35

0.2

1.22

0.87

1.34

1.16

1.7

0.3

0.18

1.26

0.86

1.41

1.11

1.55

2.03

0.32

0.21

1.14

0.92

1.29

1.2

1.59

1.96

0.3

0.19

1.13

0.88

1.29

1.13

1.54

0.29

1.95

0.29

0.17

1.14

0.86

1.3

1.07

1.51

0.64

0.3

1.97

0.35

0.1

1.07

0.85

0.61

0.26

2.02

0.3

0.13

1.2

0.76

0.32

1.17 1.31

1.12

1.73

0.93

1.57

Discussion/Conclusion: Obtained DKI and DTI values along several tracts can be useful in studies of patients with WM lesions and STBI as reference ones. References: 1. E. Pogosbekian et al. ESMRMB 2015 Congress Book of Abstracts 3 s308. 2. F. Tovar-Moll et al. Proceedings of the National Academy of Sciences, 111(21), pp. 7843–7848. 3. J. Leon-Carrion et al. Brain research, 1476, pp. 22–30. 4. E. Fieremans et al. AJNR Am J Neuroradiol 34:2105–12 (2013). 5. T. Dhollander et al. ISMRM Workshop on Breaking the Barriers of Diffusion MRI, 2016, 5.

Neuroimaging - Preclinical 661 Intrinsic ignition describes functional brain alterations in a rat model of Alzheimer’s Disease with and without neurocognitive stimulation R. Nitsche1, A. Sanjuan Tomas2, E. Mun˜oz-Moreno3, G. Soria3, R. Tudela4, G. Deco2 1 Interdepartmental Centre for Mind/Brain Sciences (CIMeC), University of Trento, Trento/ITALY, 2Computational Neuroscience Group, Center for Brain and Cognition, Department of Information and Communication Technologies, Universitat Pompeu Fabra, Barcelona/SPAIN, 3Experimental 7T MRI Unit, Institut d’Investigacions Biome`diques August Pi i Sunyer (IDIBAPS), Barcelona/SPAIN, 4GIB-UB, Centro de Investigacio´n Biome´dica en Red en Bioingenierı´a, Biomateriales y Nanomedicina (CIBER-BBN), Barcelona/SPAIN Purpose/Introduction: Animal models of Alzheimer’s Disease (AD) allow a longitudinal follow-up of the disease and testing the effectiveness of therapies. In this study, a transgenic rat model, Tg344-AD [1] is used to evaluate how brain function is altered by the disease and what the effect of neurocognitive stimulation is. An intrinsic ignition framework is considered to analyse resting state functional magnetic resonance images (rs-fMRI) to explain brain activity [4]. Subjects and Methods: rs-fMRI was longitudinally acquired in control Fisher (n = 10) and Tg344-AD (n = 10) rats performing nonmatching-two-sample (DNMS) task [3] between 5 and 15 months of age. Four additional Tg344-AD rats without neurocognitive training were scanned at similar time points. The overall datasets included 23 rs-fMRI scans of control rats, 35 of transgenic rats performing DNMS (AD-DNMS) and 11 without DNMS (AD). Each subject brain was parcellated using an adaptation of the atlas in [4]. From the regional time-series, the Ignition-Driven Mean Integration (IDMI), measuring the regions ability to spread information across the brain, as well as Ignition-Driven Variability (IDV), showing its variation over time, were computed. They were averaged over subjects for every region (rIDMI and rIDV). Additionally, the standard deviation of the IDMI across regions was analysed as a measure for the strength of hierarchical organisation between regions. Those measures were compared between groups using Kruskal– Wallis Tests and post hoc comparisons (Tukey HSD). The IDMI and IDV were compared analogously for every region and controlled using the Benjamin-Hochberg procedure. Results: Compared to the control group, rIDMI was significantly lower for the AD (p \ 0.001) and AD-DNMS (p \ 0.05) groups (Figure 1). Significant differences between all group pairs were observed for rIDV as well as for mean IDMI and IDV values (p \ 0.001). The hierarchy measurement differed significantly between AD and control rats (p \ 0.01, Figure 2). Furthermore, IDMI and IDV values of the control group were higher than both AD groups in 49 and 43 regions respectively. For IDMI, 8 regions were significantly different between groups (p \ 0.05, Figure 3), namely the right amygdala (rA), piriform (rPFC), and retrosplenial cortex (rRSC), the right and left visual cortex (rVC, lVC), and the left cingulate (lCC) and orbitofrontal cortex (lOFC).

123

S622

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 the alterations in the Tg344-AD rat model with and without DNMS sessions. Also we showed eight regions where they are maximal, including the lRSC and lPFC, which both have previously been found to be affected by early AD [5–9]. References: [1] Cohen, R. M., Rezai-Zadeh, K., Weitz, T. M., Rentsendorj, A., Gate, D., Spivak, I., et al. (2013). A transgenic Alzheimer rat with plaques, tau pathology, behavioral impairment, oligomeric Ab and frank neuronal loss. J. Neurosci. 33, 6245–6256. [2] Deco, G., Kringelbach, M. L. (2017) Diversity of computation: using a novel ‘intrinsic ignition’ framework to study brain communication. Under review. [3] Callaghan, C. K., Hok, V., Della-Chiesa, A., Virley, D. J., Upton, N., & O’Mara, S. M. (2012). Age-related declines in delayed nonmatch-to-sample performance (DNMS) are reversed by the novel 5HT6 receptor antagonist SB742457. Neuropharmacology, 63 (5), 890–7. [4] Schwarz AJ, Danckaert A, Reese T, Gozzi A, Paxinos G, Watson C, Merlo-Pich EV, Bifone A. A stereotaxic MRI template set for the rat brain with tissue class distribution maps and co-registered anatomical atlas: application to pharmacological MRI. Neuroimage 2006 32(2): 538. [5] Reyes, P. F., Golden, G. T., Fagel, P. L., et al. (1987) The Prepiriform Cortex in Dementia of the Alzheimer Type. Arch Neurol. 44(6):644–645. [6] Li, W., Howard, J. D., Gottfried, J. A., (2010) Disruption of odour quality coding in piriform cortex mediates olfactory deficits in Alzheimer’s disease. Brain 133; 2714–2726. [7] Saiz-Sanchez, D., De la Rosa-Prieto, C., Ubeda-Banon, I., Martinez-Marcos, A. (2015) Interneurons, tau and amyloid-b in the piriform cortex in Alzheimer’s disease. Brain Struct Funct. 220(4):2011–25. [8] Desgranges, B., Baron, J.-C., Laleve´e, C., Giffard, B., Viader, F., de la Sayette, V., Eustache, F. (2002) The neural substrates of episodic memory impairment in Alzheimer’s disease as revealed by FDG–PET: relationship to degree of deterioration. Brain 125 (5): 1116–1124. [9] Nestor, P. J., Fryer, T. D., Ikeda, M., Hodges, J. R. (2003) Retrosplenial cortex (BA 29/30) hypometabolism in mild cognitive impairment (prodromal Alzheimer’s disease). Eur. J. Neurosci. 18, 2663–2667.

662 Optimization of ultrahigh field micro-MRI methods to monitor brain disorders in Zebrafish model of depression U. Roy1, M. Schaaf2, J. Matysik1, A. Alia3 1 Faculty of chemistry and mineralogy, Institute for analytic chemistry, Leipzig/GERMANY, 2University of Leiden, Institute of Biology Leiden, Leiden/NETHERLANDS, 3University of Leipzig, Institut fu¨r Medizinische Physik und Biophysik, Leipzig/GERMANY

Discussion/Conclusion: The Intrinsic Ignition measures can explain

123

Purpose/Introduction: Glucocorticoid (GC) resistance syndrome is stress related dementia caused from the dysfunctional hypothalamic– pituitary–adrenal (HPA) axis. It is characterized by diminished cortisol signalling through GC receptor (GR). However, it is not clear whether the GR deficiency can cause any specific structural changes in the brain. Zebrafish is increasingly used as model organism for understanding brain diseases including neurodegenerative disorders. However due to small size of the brain, non-invasive imaging methods in zebrafish are challenging. In this study, we have optimized and applied high resolution lMRI methods to examine anatomical

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 details in the brain of GR deficient (gr-/-) zebrafish (1) to identify changes which could be attributed to the depression. Subjects and Methods: MRI measurements were performed at 9.4 T vertical bore system (2), using a birdcage radiofrequency coil with an inner diameter of 10 mm and a 1 Tm-1 gradient insert (Bruker Analytic, Germany). T2-weighted MR images were acquired by RARE sequence [echo time (TE) = 10.567 ms (22.45 ms effective), repetition time (TR) = 5 s, averages = 64, echo train length = 4, FOV = 1.22 cm2, matrix = 2562, in plane resolution 47 lm]. For T2 relaxation time measurement, MSME sequence was used. The wholebody fat mapping was done using CHESS sequence [TR = 1 s, TE = 3.4 ms, FOV = 22 cm2 and matrix = 1282]. 1H chemical shift imaging (CSI) was performed at 7.0 T magnet with TE = 15 ms, TR = 1.5 s and NS = 256. Signals between 0.80 and 1.25 ppm were chosen to map fat signals.

Results: Optimized RARE sequence provided sufficient resolution to identify specific markers in both sagittal and coronal direction from adult zebrafish brains. Some structural differences were seen in the two groups which may be linked to depression-like behaviour. CSI and CHESS measurements map fat in the control and GR mutant fish. Signals between 0.80 and 1.25 ppm were chosen to map fat signals. Discussion/Conclusion: In spite of the small size of the brain, we have characterised anatomical details of adult Zebrafish brains which suggested that glucocorticoid signalling is responsible for hypertrophy of some of the brain structures. Optimization of CSI and CHESS sequences adds new toolbox for studying the localization of metabolites in vivo in the pathology of dementia. Furthermore, this study demonstrates that lMRI is an efficient tool to successfully visualize human diseases in zebrafish models. References: (1) Ziv L et al. (2013) Mol Psychiatry 18: 681. (2) Kabli et al. (2010) Zebrafish 7: 143.

663 Evaluation of texture features on resting-state networks of a rat model of alcohol use disorders ´ . Pe´rez-Ramı´rez1, A. Dı´azR. Ortiz-Ramo´n1, S. Ruiz-Espan˜a1, U Parra1, R. Ciccocioppo2, S. Canals3, D. Moratal1 1 Center for Biomaterials and Tissue Engineering, Universitat polite`cnica de Vale`ncia, Valencia/SPAIN, 2School of Pharmacy, University of Camerino, Camerino/ITALY, 3Instituto de Neurociencias, Consejo Superior de Investigaciones Cientı´ficas, Univerdidad Miguel Herna´ndez, Alacant/SPAIN

S623 AUD are associated to alterations in functional connectivity in brain networks. In this study, our objective was to explore a novel source of biomarkers for identifying subjects with AUD by performing a texture analysis applied to resting-state functional MRI (rs-fMRI) data acquired in a rat model of AUD. Subjects and Methods: Eighteen Marchigian Sardinian alcoholpreferring rats were used in a longitudinal study. Two time points were considered: before alcohol consumption (control condition) and after 30 days of alcohol drinking. T2*-weighted rs-fMRI brain images were acquired with a Bruker Biospec 7T scanner (resolution = 0.26 9 0.26 9 1 mm3). The rs-fMRI images were preprocessed using FSL 5.07. First, images were motion corrected and then, they were registered to a T2-weighted image template. Group independent component analyses (ICA) considering control and alcohol conditions were performed and nine independent components (ICs, spatial maps with minimal redundancy) were identified as resting-state networks (RSNs) [1]. Binary masks considering the positive weights of the ICs were extracted to focus the texture analysis on these networks (Fig. 1). To compute the texture features, the MATLAB toolbox Radiomics [2] was used. Twenty-two rotation-invariant features were extracted from two texture methods: 9 features from the gray-level co-occurrence matrix (GLCM) and 13 from the gray-level run-length matrix (GLRLM). To evaluate the statistical significance of the features, the p-value provided by the Wilcoxon signed rank test was used. The pvalues were adjusted with the Holm method to address the problem of multiple comparisons. Results: Twenty features from five ICs showed statistical significance (p \ 0.05), and 6 of them provided a p \ 0.01 (Fig. 2), which is a satisfactory result given the conservative adjustment applied. The box plots also indicated that control and alcohol conditions could be discriminated using these features (Fig. 3). It is noteworthy that four ICs did not provide any significant feature and, on the contrary, nine of the 20 significant features derived from the same IC (striatal network). Discussion/Conclusion: Texture analysis over RSNs extracted from rs-fMRI brain images in a rat model of AUD provides features that could be used as biomarkers to discriminate between alcohol and control conditions. Furthermore, it can help identify brain networks most affected by (or better defining a) pathological state, as it is the case for the striatum in the presented AUD model. References: [1] U. Pe´rez-Ramı´rez et al., ‘‘Brain functional connectivity alterations in a rat model of excessive alcohol drinking: a resting-state network analysis,’’ In: The 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Conference (EMBC 2017), Jeju, Korea, 2017. [2] M. Vallie`res et al., ‘‘A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metastases in softtissue sarcomas of the extremities,’’ Phys. Med. Biol., vol. 60, pp. 5471–96, 2015.

Purpose/Introduction: Alcohol use disorders (AUD) are a major risk for disease and disability. Neuroimaging analysis has revealed that

123

S624

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S625

664 Registration of mouse brain microscopy images to a MR mouse brain atlas for locating interneuron cells: a preliminary study ´ . Pe´rez-Ramı´rez1, A. Dı´az-Parra1, O. R. Ortiz-Ramo´n1, A. Llorca2, U Marı´n2, D. Moratal1 1 Center for Biomaterials and Tissue Engineering, Universitat polite`cnica de Vale`ncia, Valencia/SPAIN, 2Department for Developmental Neurobiology, King’s College London, London/ UNITED KINGDOM Purpose/Introduction: Mouse brains are frequently studied using high resolution fluorescence microscopy images to locate interneuron cells, although it is difficult to locate the cells in a specific brain region. The objective of this work is to register the microscopy images of a mouse brain with a MR mouse brain atlas to automatically identify the position of the interneuron cells within the brain and the anatomical region where they are located. Subjects and Methods: The microscopy images of the mouse brain selected in this study were originally grouped into six mosaics of DAPI-stained coronal brain slices (Fig. 1a), with an in-plane resolution of 6.05 lm and a separation between slices of 100 lm. All the slices were automatically segmented and realigned to reconstruct a 3D brain image. To segment the slices, the Otsu’s method was used in combination with morphological reconstruction algorithms. To realign the slices, an algorithm that combined a method based on the Dice Similarity Coefficient with a rigid transformation (translation and rotation) was used. The processing of the microscopy images was performed using a software developed in MATLAB. The 3D MR mouse brain atlas used in this work was segmented into 62 structures using a template obtained by normalizing, registering and averaging 3D T2-weighted MR images of 40 adult C57Bl/6 J mice into one volume with an isotropic resolution of 32 lm [1]. The images involved in the registration process are shown in Fig. 1b. The 3D microscopy image was registered with the averaged MR template using a two-stage procedure implemented in MATLAB. First, a 12 parameter affine transformation (translation, rotation, zoom and shear) was applied to the template to account for the major differences in shape and position. Afterward, a non-linear transformation was applied to take care of the smaller-scale differences in brain anatomy. Finally, the same transformations were applied to the atlas. The registration process was implemented in MATLAB using the suite SPM. Results: The results showed that the registered atlas fitted properly to the 3D microscopy image, mainly in the central coronal slices, and the cerebral regions could be identified in the microscopy images, as shown in Fig. 1c. However, further research should be done to achieve a more accurate registration. Discussion/Conclusion: These preliminary results showed the potential of this MR mouse brain atlas to serve as a reference to label the brain areas in a microscopy image. With this approach, interneurons cells could be located precisely in mouse brain microscopy images.

References: [1] A. E. Dorr et al., ‘‘High resolution three-dimensional brain atlas using an average magnetic resonance image of 40 adult C57Bl/6J mice,’’ Neuroimage, vol. 42, no. 1, pp. 60–69, 2008.

123

S626

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

New Contrasts 665 Optimization of imaging parameters for fast bound pool fraction estimation from a single off-resonance magnetization transfer measurement M. Soellradl, L. Pirpamer, F. Fazekas, C. Langkammer, S. Ropele Department of Neurology, Medical University of Graz, Graz/ AUSTRIA Purpose/Introduction: The bound pool fraction f, a quantity representing the relative density of protons bound to macromolecules, has received interest as biomarker for the assessment of myelin content in brain tissue. For the fast estimation of f, Yarnykh proposed to constrain the model parameters to determine f from a single offresonance magnetization transfer (MT) measurement1. In the original work, the error of this method was simulated for varying MT-pulse amplitude FAMT and off-resonance frequency D at a constant flip angle a and repetition rate TR. Here, we present an improved new numerical error-model that takes a and TR into account. Subjects and Methods: For the proposed error model shown in Figure 1 the error di is estimated after addition of noise with standard deviation rnoise, defined in Equation (1), to the normalized signal mz of the pulsed MT model 2.

The signal to noise ratio of the reference image SNRref is modeled as constant value SNRmax scaled with the signal equation of the spoiled gradient echo sequence where T1obs is the observable relaxation time. To study the impact of the imaging sequence parameters a synthetic data set was created with N = 1000 normally distributed model parameters according to white matter (WM) values found in literature 1 . (Simulated sequence parameters: TR = 40–200 ms/a = 5–90/ FAMT = 400–1000/D = 3–10 kHz/SNRmax = 500–1000).

Results: The results in Figure 2 indicate a strong dependence of the error from a and TR with the minimal error shifted towards higher a for larger TR with the biggest dynamic range for TR \ 60 ms. Moreover the importance of SNRmax is shown by the decreasing error from SNRmax = 500 to SNRmax = 1000. By limiting the analyses to TR \ 60 ms we found the minimal error for FAMT between 600 and 700 with D ranging from 3.8 to 5.2 kHz.

123

Discussion/Conclusion: Optimal D and FAMT for minimizing the error of the fast f estimation are in line with previous reported results1. Furthermore, the simulations show that TR and a strongly influence the error of the method with a possible 30% reduction when TR is increased from 40 ms i.e. to 60 ms. In conclusion, extending the MT-pulse parameter (FAMT = 600900, D = 4–7 kHz) recommended in 1, our simulations suggest a = 20 and TR between 50 and 60 ms for accurate and fast bound pool fraction estimation in WM. References: 1. Yarnykh VL. Fast macromolecular proton fraction mapping from a single off-resonance magnetization transfer measurement. Magn Reson Med. 2012 Jul;68(1):166–78. 2. Yarnykh VL., Yuan C. Cross-relaxation imaging reveals detailed anatomy of white matter fiber tracts in the human brain. Neuroimage. 2004 Sep;23(1):409–24.

666 ZTE imaging of the mouse knee in vivo at 9.4 T R. In ‘T Zandt Lund Bioimaging Center LBIC, Lund University, Lund/SWEDEN Purpose/Introduction: Zero Echo Time Imaging (ZTE) is a 3D imaging technique for imaging of samples with short transverse relaxation times [(1) and references therein]. The image contrast is like a proton density image. The sequence is virtually silent but requires high RF power levels. It is our aim to use the ZTE sequence for high resolution imaging in vivo, and as an example, to visualise the patellar ligament in the mouse knee in vivo. Subjects and Methods: The ZTE sequence was tested in vivo using a mouse brain cryo-coil connected to a Bruker Avance III 9.4 T. We used the highest possible bandwidth of 400 kHz while maximizing the gradient usage (660 mT/m, duty cycle 93%, gradient strength 39.90%) at a FOV of 35 9 35 9 35 mm3 and a matrix size of 512 9 512 9 512. Off-line reconstruction was chosen to allow for full polar sampling (825130 projections). The RF pulse length was 1 ls and for a 0.88 pulse, 4.9 W peak power was needed. At a TR = 2 ms, the total scan time was 28 min. For comparison, a 3DFLASH with TE = 3.45 ms/TR = 18 ms with simular scan time (37 min) and identical pixel resolution (68 lm) was acquired. Results: The in vivo ZTE image of the mouse knee reveals the patellar ligament (figure 3 left, red arrows) while in the 3D-flash image the ligament is black.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S627

667 SE-WMRI Fast Spin Echo Imaging E. Wu, P.-W. Cheng, T.-D. Chiueh, J.-H. Chen Electrical Engineering, National Taiwan University, Taipei/TAIWAN Purpose/Introduction: Fast Spin Echo (FSE) is one of the most common imaging methods among MRI due to its fine contrast and signal strength, its scan time is reduced by applying a [1 Echo Train Length (ETL) to acquire multiple k-lines in one excitation. However, greater ETL causes image quality degradation and builds up specific absorption rate (SAR) in the subject as well. In this study, we apply SEWMRI to acquire FSE images with without the drawbacks of high ETL. Subjects and Methods: SE-WMRI uses a specialized zig-zag sampling trajectory and reduced phase encoding steps to acquire images of identical quality, the acceleration rate is defined as Wideband factor ‘‘W’’ [1]. In this study, a W = 2, S = 5 (5-segmented) SEWMRI Bruker FSE sequence was programmed based on its gradient echo counterpart. The in vivo images were taken on a Biospec 94/20 system using surface head coil, with/without SE-WMRI applied. The standard images all share the following parameters: FOV = 2.4 9 2.4 cm2, 1 mm thickness, Matrix size = 256 9 256, TR/TE = 3000/14.16 ms, NEX = 2. Scan time for standard FSE 8 and 4 are 6m24 s and 12m48 s respectively. Results: Applying SE-WMRI directly reduces the scan time in half, using 6m24 s instead of 12m48 s in ETL = 4 imaging while SNR was not sacrificed (shown in Fig. 1). on the other hand, increasing ETL twofold in standard imaging also halves the scan time. Comparing the ETL = 4, W = 2 image with ETL = 8 image, we can clearly see the SNR advantage of the former setting (Fig. 2).

Discussion/Conclusion: Using ZTE, imaging of the patellar ligament was feasible of the mouse knee in vivo. The pixel resolution achieved was 68 lm isotropically. A decrease in the bandwidth will allow for a smaller FOV but this will likely cause loss of short T2 components. At lower bandwidth, the minimum TR will increase and the scan time approaches 1 h. FOV should be such that all signal picked up by the RF-coil is within the FOV to avoid artefacts. This limits the minimal FOV. The total current scan time of 30 min is reasonable for in vivo scanning but the reconstruction stage is performed off-line to allow for a fully polar sampled k-space and the reconstruction will take more than 1 h on our Bruker console with 64 GB memory. In summary, the ZTE approach is stable but its application should be chosen with care due to the lack of high contrast, its high RF power levels needed and its spatial resolution mainly depending on the gradient performance and not us such the RF-coil used [see for an excellent overview (1)]. More work to better understand the ZTE contrast for pathology is ongoing. References: 1. Weiger and Pruessmann eMagRes, 2012, Vol 1: 311–322. DOI 10.1002/9780470034590.emrstm1292 .

123

S628

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: Other than improved image quality or scan time reduction, another benefit of SE-WMRI is the reduced SAR values, this could provide a relieve in clinical practices where safety is a significant issue. The difference in image SNR could also be translated into higher spatial resolution or less number of averages, providing more options for users. Since SE-WMRI uses switching gradients, it should be sensitive to non-uniform gradients and magnetic susceptibilities. Regions that contain tissue-air interface or blood vessels seems to slightly resemble gradient echo contrasts. In summary, SE-WMRI FSE imaging shows great promise to either shorten the scan time or increase SNR due to shorter TE. References: 1. Wu, E.L., et al., Single-frequency excitation wideband MRI (SEWMRI). Med Phys, 2015. 42(7): p. 4320–8.

668 Improved signal-to-noise ratio with highly asymmetric spin-echo EPI (HASE-EPI) M. Shrestha, U. No¨th, R. Deichmann Brain Imaging Center (BIC), Goethe University Frankfurt, Frankfurt/ Main/GERMANY Purpose/Introduction: Structural and diffusion-weighted imaging (SI, DWI) in biomedical research require high spatial resolution and signal-to-noise ratio (SNR). Techniques frequently employ EPI [1] with spin-echo (SE) preparation [2], adding diffusion-weighting gradients (DWG) for DWI [3]. In general, the SE coincides with central k-space sampling during the EPI-readout. Here, the ‘‘highly asymmetric spin-echo’’ (HASE) concept is proposed, where the SE occurs prior to the EPI-readout (see Fig. 1), yielding T2-weighting during SE preparation and T*2-weighting during the EPI-readout. Simulations and experiments show that HASE may yield increased SNR. Subjects and Methods: Sequence: Fig. 1 shows the HASE module for SI (a) and DWI (b). The SE forms at TE1. During the matrix-sizedependent TE2, there is a T*2-decay, yielding the signal loss exp(-TE2/T*2). However, conventional SE-EPI would require insertion of an additional TE2 within s1 to achieve SE formation at readout centre, causing the total TE2-dependent signal loss exp(-2TE2/T2). Thus, HASE yields higher SNR for TE2/T*2 \ 2TE2/T2 or: T*2 [ T2/2. Simulation: Signal gains of HASE-EPI versus SE-EPI were calculated for different sampling schemes affecting the EPI-readout duration: matrix-size = [64, 96, 128], echo-spacing = [0.54, 0.82, 0.86] ms, full EPI-readout or reduced EPI-readout via twofold acceleration (iPAT = 2) with/without 6/8 partial-Fourier. For white and gray matter, T2/T*2 = 60/50 and 65/45 ms were assumed, respectively. Protocols: Two DWI (in-plane-resolutions: 2 mm/1.5 mm) and three SI protocols (in-plane-resolutions: 3 mm/2 mm/1.5 mm) were used with full or reduced (iPAT = 2) EPI-readout, FOV = 192 9 192 mm2, 60 axial slices (slice-thickness = 2 mm, no gap, in-plane-resolution = [3 9 3.2 9 2.1.5 9 1.5]mm2 with corresponding matrix-size = [64, 96, 128], TE1 = 14/72 ms (SI/DWI), TR/volume = [5, 9, 12] s for SI and [12, 15] s for DWI. Experiments: Healthy volunteers were scanned on a 3T whole-body MRI-scanner, using an 8-channel phased-array head-RX-coil and the above protocols.

123

Results: Simulation: Fig. 2 shows that HASE-EPI yields higher SNR than standard SE-EPI, especially for higher matrix-sizes and full EPIreadout. Experiments: Fig. 3 shows in vivo results. In SI, HASE-EPI yielded signal gains in corpus-callosum/caudate-nucleus of [15–41]%/ [15–49]% (Fig. 3a, full EPI-readout) and [1–22]%/[2–19]% (Fig. 3b, iPAT = 2), for in-plane resolutions = [3 9 3–1.5 9 1.5] mm2. In DWI, gains in corpus-callosum were [12–17]% (Fig. 3c, full EPIreadout) and [1–14]% (Fig. 3d, iPAT = 2) for in-plane resolutions = [2 9 2–1.5 9 1.5] mm2. Signal gains in SI and DWI differ mainly due to the TE1-difference.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: The proposed HASE-EPI yields an EPIreadout-independent TE1, thus allowing for higher SNR. Differences in predicted (simulation) and measured (in vivo) signal gains may be due to imperfect refocusing pulses and deviations of actual from assumed T2/T*2 values. Signal gains were lower in DWI but may be increased by shortening TE1, e.g. via using stronger DWG or acquiring navigator-echoes separately. Signal gains may increase at high magnetic-field strengths, due to shortened T2 and T*2. References: [1] Mansfield, P. (1977). J. Phys. C: Solid State Phys., 10(3), L55. [2] Hahn, E.L. (1950). Phys. Rev., 80(4), 580. [3] Stejskal, E.O., & Tanner, J.E. (1965). J. Phys. Chem., 42(1), 288–292.

669 Adaptive Wavelet Thresholding for Profile-Encoding Reconstruction of Balanced Steady-State Free Precession Acquisitions

S629 Purpose/Introduction: A powerful strategy to suppress banding artifacts while maintaining scan efficiency in balanced steady-state free precession (bSSFP) imaging is to accelerate multiple phase-cycled acquisitions [1, 2]. We have recently proposed a profile-encoding framework (PE-SSFP) where undersampled bSSFP acquisitions are jointly reconstructed across phase-cycles to recover unacquired data. The PE method was shown to achieve superior image quality compared to conventional parallel imaging and compressed sensing reconstructions [2]. PE enforces sparsity by soft-thresholding of the entire wavelet tree via a single, manually determined threshold. Here we propose an improved method where the optimal wavelet threshold for each sub-band is automatically determined from the data. The proposed method reduces reconstruction artifacts while preserving detailed tissue depiction in bSSFP images. Subjects and Methods: In multiple-acquisition bSSFP imaging, several images with different phase-cycling increments are acquired. Because the spatial locations of the banding artifacts differ across phase cycles, individual images can be combined to suppress artifacts. In a recent study [2], we developed a PE reconstruction that linearly synthesizes unacquired data in accelerated phase-cycled acquisitions, and leverages joint sparsity and total variation penalties to suppress aliasing and noise. As input, the PE method requires manual specification of a single wavelet threshold that is enforced on the entire wavelet tree. Here, we propose to automatically determine appropriate thresholds for each wavelet subband by the PES-L1 method, which enforces sparsity by projecting wavelet coefficients onto the epigraph set of the L1-ball. This epigraph set represents the family of L1-balls with the maximum size equal to the L1-norm of the wavelet coefficients prior to projection. A unique projection that yields wavelet coefficients closest to the L1-ball of size 0 is then determined [3, 4]. The proposed method was demonstrated on brain phantom and in vivo brain acquisitions. PE reconstructions were obtained using PES-L1, and conventional soft-thresholding (with matched average threshold). Image quality was assessed via the peak signal-to-noise ratio (PSNR). Results: Figure 1 shows reconstructions of four- and eightfold phantom acquisitions; Figure 2 shows reconstructions of eightfold accelerated in vivo acquisitions. The proposed method achieves more detailed tissue depiction compared to conventional soft-thresholding. This assessment is supported by PSNR measurements listed in Table 1.

M. Shahdloo1, E. Ilıcak1, M. Tofighi2, E.U. Saritas1, A.E. C¸etin3, T. C¸ukur4 1 Department of Electrical and Electronics Engineering and National Magnetic Resonance Research Center (UMRAM), Bilkent University, Ankara/TURKEY, 2Department of Electrical Engineering, Pennsylvania State University, PA/UNITED STATES OF AMERICA, 3 Department of Electrical and Electronics Engineering, and IL, US, University of Illinois at Chicago, Department of Electrical and Computer Engineering, Bilkent University, Ankara/TURKEY, 4 Department of Electrical and Electronics Engineering, National Magnetic Resonance Research Center (UMRAM), and Neuroscience Program, Graduate School of Engineering and Science, Bilkent University, Ankara/TURKEY

123

S630

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion/Conclusion: The proposed method enforces sparsity via an adaptive threshold in each wavelet subband to improve sensitivity for sparse recovery. The thresholds are determined from the acquired data without supervision. As a result, the proposed method reduces noise/aliasing artifacts and improves spatial resolution compared to reconstructions based on conventional wavelet thresholding. References: C¸ukur T., Accelerated Phase-Cycled SSFP Imaging With Compressed Sensing. IEEE Transactions on Medical Imaging 2015. doi: 10.1109/TMI.2014.2346814.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S631

Ilicak E., Senel L.K., Biyik E., C¸ukur T., Profile-encoding reconstruction for multiple-acquisition balanced steady-state free precession imaging. Magnetic Resonance in Medicine 2016. doi: 10.1002/mrm.26507. C¸etin A. E., Tofighi M., Projection-Based Wavelet Denoising [Lecture Notes]. IEEE Signal Processing Magazine, vol. 32, no. 5, pp. 120–124, 2015. Tofighi M., Yorulmaz O., Ko¨se K., Yıldırım D. C., C¸etin-Atalay R., C¸etin A. E., Phase and TV based convex sets for blind deconvolution of microscopic images. IEEE Journal of Selected Topics in Signal Processing 2016, 10(1), 81–91.

670 Lipid nanocapsules as a prognostic tool in cancer J. Nel1, B. Gallez1, F. Franconi2, N. Joudiou1, L. Lemaire3 1 Biomedical Magnetic Resonance Unit (REMA), Universite´ catholique de Louvain, Woluwe-Saint-Lambert, Brussels/BELGIUM, 2 PRISM, IBS-CHU, Universite´ Angers, Angers/FRANCE, 3Micro et Nanomedecines translationnelles (MINT), INSERM 1066, CNRS 6021, Universite´ Angers, Angers/FRANCE Purpose/Introduction: Hypoxia is a characteristic of solid tumours, wherein gradients of O2 arise within the tumour mass due to its abnormal vasculature being unable to meet the O2 demand as it grows. This leads to more aggressive tumours which are resistant to radiation and cytotoxic chemotherapy [1]. For tumor hypoxia determination, MR techniques are of interest. Changes in tissue O2 concentrations produce changes in the relaxation rate R1 (=1/T1) of water, thus T1 demonstrates a sensitivity to dissolved O2 (which acts as a T1-shortening paramagnetic probe). An O2-induced increase in the R1 has the potential to provide measurements in fluctuations in the O2 level of tissue. Jordan et al. [2] describes an MR method allowing for the rapid mapping of changes in tissue oxygenation based on the higher solubility of O2 in lipids than in water (MOBILE). The aim of this study was to evaluate hypoxia by measuring O2 levels in tumours after lipid introduction in the form of lipid nanocapsules (LNCs). Subjects and Methods: LNCs were synthesised via the PIT method [3] and size and polydispersity (PDI) was determined by DLS analysis. Prior to in vivo experimentation, the ability of the LNCs to measure pO2 was assessed in vitro using distinct samples with differing O2 levels, namely 0, 21, and 100% (corresponding to 0, 0.2, and 1 atm, respectively) and the change in lipid T1 was linearly correlated with the respective pO2 level using MOBILE. Furthermore, a carbogen challenge was used to induce changes in tissue pO2. To study hypoxia, a brain tumour model was utilised with osmotically adjusted LNCs for in vivo evaluation. Results: LNCs were synthesised with an average hydrodynamic size of 53.9 nm ± 0.7 and a PDI of 0.04 ± 0.01. For in vivo validation, the osmotically adjusted LNC preparation was injected either into the tumour site or into the normal brain. T1 relaxation was determined by drawing a ROI around the lipid hypersignal and analysing the T1 relaxation rates pixel-by-pixel using MatLabTM software to enable both a mean T1 value and a mapping of the tumour pO2. T1 relaxation rates were mapped before and after the O2 challenge to better test the capability of the LNCs to facilitate a T1 relaxation response. Discussion/Conclusion: Preliminary results indicate that tumor pO2 can be mapped by the MOBILE MR sequence after a single introduction of LNCs. The lipid T1 map was sensitive to variations in oxygenation induced by an oxygen challenge.

References: [1] Ramachandran S, Ient J, Gottgens E-L, Krieg AJ, Hammond EM. 2015 Epigenetic Therapy for Solid Tumours: Highlighting the Impact of Tumour Hypoxia. Genes. 6(4), 935–956. [2] Jordan BF, Magat J, Colliez F, Ozel E, Fruytier AC, Marchand V, Mignion L, Bouzin C, Cani PD, Vandeputte C, Feron O, Delzenne N, Himmelreich U, Denolin V, Duprez T, Gallez B. 2013. Mapping of oxygen by imaging lipids relaxation enhancement: a potential sensitive endogenous MRI contrast to map variations in tissue oxygenation. Magnetic Resonance in Medicine. 70, 732–744. [3] Heurtault B, Saulnier P, Pech B, Proust JE, Benoit JP. 2002. A novel phase inversion-based process for the preparation or lipid nanocarriers. Pharmaceutical Research. 19, 875–880.

671 Study of suitability of SPRITE pulse sequence for dental MRI H. Sa´nchez-Izquierdo1, E. Diaz-Caballero2, A. Nacev3, D. GrauRuiz1, J.P. Rigla2, R. Hilaman3, J.M. Gonzalez1, G. Puchalt1, J.M. Benlloch1 1 Instituto de Instrumentacio´n para Imagen Molecular (I3 M), Universitat Polite`cnica de Vale`ncia, Valencia/SPAIN, 2Research and development, Tesoro Imaging S.L., Alicante/SPAIN, 3Reasearch and Development, Weinberg Medical Physics, Rockville/UNITED STATES OF AMERICA Purpose/Introduction: MRI has been traditionally used to study soft tissue. Currently, the most used technique to obtain teeth images is projection radiography or cone beam computer tomography (CBCT) [1]. The main disadvantage of these methods is that they involve ionizing radiation. Typical MRI pulse sequences are not appropriate to catch signal from short T2* tissues, such as bones or teeth, because their relaxation times are of the order of the gradient rise time. However, there are pulse sequences, such as UTE, ZTE [2] or SPRITE [3], which seem more suitable for these kinds of tissues. In this work, we have implemented SPRITE pulse sequences in order to study the advantages they can provide for obtaining teeth images compared to typical pulse sequences, such as Gradient Echo [4].

123

S632 Subjects and Methods: SPRITE (Single-Point Ramped Imaging with T1 Enhancement) sequences allow visualizing tissues with T2* below 100 ls, so they are suitable for teeth imaging (T2* of teeth is between 12 ls–1 ms for dentin [5] and 4–240 ls for enamel [6]). This is possible since the RF pulse is applied at the same time that gradients, avoiding having to turn on the gradients after every RF pulse. Dental samples were examined ex vivo using a low-cost tabletop MRI with a low magnetic field (B0 = 0.33T) [7]. All the pulse sequences programming, as well as the system control, has been implemented in MATLAB [8] and C/C++. Results: Images of a premolar tooth were obtained using Gradient Echo and SPRITE sequences. Several essays were performed in order to find the proper pulse parameters for each sequence. Fig. 1.b shows the results of an 80 9 80 Gradient Echo image (50 scans, TE = 223.43 ls and TR = 100.48 ms) whose acquisition required 7 min. Fig. 1.c shows a 100 9 100 SPRITE image (1 scan, TE = 80.46 ls and TR = 427.27 ms) which required 40 min. Signal-to-noise ratio (SNR) obtained for both images were, respectively, 0.9286 and 3.045 [9].

Discussion/Conclusion: SPRITE sequences provide better SNR with much less averaged scans, so they are an interesting option to image teeth since short T2* problems have less influence. Another advantage of this kind of sequence is that it is much more silent than Gradient Echo due to the use of ramped gradients. Its main drawback is the long acquisition time, so parallel imaging or fast k-space sampling (spiral, conical, etc.) should be implemented in order to reduce it [4][10]. References: [1] Durack, C., & Patel, S. (2012). Cone beam computed tomography in endodontics. Brazilian dental journal, 23(3), 179–191. [2] Weiger, M., Pruessmann, K. P., Bracher, A. K., Ko¨hler, S., Lehmann, V., Wolfram, U.,… & Rasche, V. (2012). High-resolution ZTE imaging of human teeth. NMR in Biomedicine, 25(10), 1144–1151. [3] Balcom, B. J., MacGregor, R. P., Beyea, S. D., Green, D. P., Armstrong, R. L., & Bremner, T. W. (1996). Single-Point Ramped Imaging withT 1 Enhancement (SPRITE). [4] Bernstein, M. A., King, K. F., & Zhou, X. J. (2004). Handbook of MRI pulse sequences. Elsevier. [5] Schreiner, L. J., Cameron, I. G., Funduk, N., Miljkovic´, L., Pintar, M. M., & Kydon, D. N. (1991). Proton NMR spin grouping and exchange in dentin. Biophysical journal, 59(3), 629–639. [6] Funduk, N., Kydon, D. W., Schreiner, L. J., Peemoeller, H., Miljkovic´, L., & Pintar, M. M. (1984). Composition and relaxation of the proton magnetization of human enamel and its contribution to the tooth NMR image. Magnetic resonance in medicine, 1(1), 66–75. [7] J.P. Rigla, D. Grau-Ruiz, E. Dı´az-Caballero, and A. Nacev et al. Tabletop MRI system development for intraoperative biopsy analysis. In 2016 IEEE NSS/Medical Imaging Conference (MIC), At Strasbourg, 2016. [8] MATLAB—MathWorks. https://www.mathworks.com/products/ matlab.html.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [9] Olaf Dietrich, Jose´ G. Raya, Scott B. Reeder, Maximilian F. Reiser, and Stefan O. Schoenberg. Measurement of Signal-to-Noise Ratios in MR Images: Influence of Multichannel Coils, Parallel Imaging, and Reconstruction Filters. Journal of Magnetic Resonance Imaging 26:375–385, 2007. [10] Halse, M., Goodyear, D. J., MacMillan, B., Szomolanyi, P., Matheson, D., & Balcom, B. J. (2003). Centric scan SPRITE magnetic resonance imaging. Journal of Magnetic Resonance, 165(2), 219–229.

672 Magnetic Resonance Probing Ensemble Dynamics in kSpace V. Herold, T. Kampf, P. Jakob Physics Department, University of Wu¨rzburg, Am Hubland/ GERMANY Purpose/Introduction: We demonstrate the use of spatial encoded magnetic resonance to quantify ensemble dynamics of dispersed microscopic particles below the spatial resolution. Subjects and Methods: By evaluating time series of single k-space data-points, k-dependent motion patterns can be revealed in short measurement time. The signal generation can be described analogously to the theory of dynamic light scattering [1]. Ensemble dynamics can be measured by examining time correlations in the k-space signal using the autocorrelation function or structure function [2]. To proof the feasibility of this new technique we have simulated the MR measurements with samples producing particle drift and brownian motion. MR experiments with sedimenting microspheres (D = 100 lm) verified the results of the simulations. A typical CINEFLASH-(fast-low-angle-shot)-sequence with N = 200 time-frames and a repetition time of TR = 3.6 ms was applied first with and then without spatial phase encoding [3]. Results: Fig. 1a shows a photography of the sedimenting glas spheres. For the limited temporal and spatial resolution (0.4 9 0.4 mm2) the reconstructed CINE-MR-images reveal no details about the particle size, distribution and dynamics as shown in Fig. 1b. When picking out the time course of a single kx-encoded data-point (without spatial phase encoding), and examining the temporal correlation in the signal, the imprint of the ensemble motion gets visible. In the case of drifting particles the motion is reflected in the signal oscillation, as shown for three different k-space-values in Fig. 1c. The corresponding spectra of the structure functions allow the quantification of the drift motion as shown in Fig. 1d [2].

Discussion/Conclusion: The mathematical description of this new method is very similar to the well known q-space encoding, it allows to quantify particle motion below the spatial resolution almost in realtime [4]. Since this new approach does not directly encode the particle displacement such as done with q-space encoding it is not limited by

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 relaxation times and covers a wide field of applications for particle motion in opaque media. References: [1] B. J. Berne and R. Pecora, Dynamic light scattering: with applications to chemistry, biology, and physics (Courier Corporation, 1976). [2] R. Cerbino and V. Trappe, Phys Rev Lett 100, 188102 (2008). [3] A. Haase, J. Frahm, D. Matthaei, W. Hanicke, and K.-D. Merboldt, Journal of Magnetic Resonance (1969) 67, 258 (1986). [4] P. T. Callaghan, Translational dynamics and magnetic resonance: principles of pulsed gradient spin-echo NMR (Oxford University Press, 2011).

S633 applications, such as tumor diagnosis. This work is ongoing in our group. References: [1] C. K. Lim, A. Singh, J. Heo, et al. Biomaterials. 34(28), 6846–52 (2013). [2] K. S. Kim, W. Park, J. Hu, et al. Biomaterials. 35(1), 337–43 (2014). [3] A. Ahmed, C. Zhang, J. Guo, et al. Macromolecular Bioscience. 15(8), 1105–14 (2015). [4] A. Riedinger, M. Pernia Leal, S. R. Deka, et al. Nano Letters. 11(8), 3136–41 (2011).

673 Manganese-Based Nanogels as pH Switches for Magnetic Resonance Imaging

674 Varying the mixing time of the double diffusion experiment: A better experimental design for pore size estimation

C. Caro1, M.L. Garcı´a-Martı´n1, M. Pernı´a Leal2 1 Nanodiagnostics, BIONAND, Andalusian Centre for Nanomedicine and Biotechnology, Ma´laga/SPAIN, 2Departamento de Quı´mica Orga´nica y Farmace´utica, Universidad de Sevilla, Sevilla/SPAIN

V. Methot, P. Ulloa, M. A. Koch Institute of medical engineering, University of Luebeck, Luebeck/ GERMANY

Purpose/Introduction: Hybrid smart MRI contrast agents (CA) can be obtained by the combination of a paramagnetic cation and an external stimuli responsive nanomaterial, such as hydrogels, micelles or nanoparticles [1–3]. We have developed pH-responsive nanogels (NGs) that can be loaded with paramagnetic ions, promoting high efficiency dual T1/T2 contrast agents for magnetic resonance pH imaging. The NG is sensitive to pH changes, such that protonation induces a change of the polymer hydration state and consequent swelling. The swollen NG allows water molecules to enter and interact with the Mn chelate, shortening the relaxation time (switch ON) and giving rise to positive or negative contrast on T1- or T2weighted MR images. Subjects and Methods: Characterization of pH-responsive NGs involved: transmission electron microscopy, dynamic light scattering, relaxivity measurements at low and high magnetic fields, inductively coupled plasma high resolution mass spectroscopy, and cell cytotoxicity by MTT. Synthesis of empty NGs was done according to the protocol published by Pellegrino and col. with minor modifications [4]. Internalization of Mn(II) cations inside the NGs was done as follows: the pH was lowered to 3 with HCl and then MnCl2 was added. The resulting solution was left in an orbital shaker for 72 h and then the pH was returned to 7 to close the NGs. Mn-NGs were purified by dialyzing against milliQ water. MR T1 and T2-weighted images of phantoms containing solutions of Mn-NGs at different pHs were acquired at 9.4 T. Results: The MR relaxation properties of the Mn-NGs have been investigated at clinical (1.5 T) and pre-clinical (9.4 T) magnetic fields. When the Mn-NGs are taken to a pH below 4, the polymeric matrix swells allowing the water molecules to interact with the manganese chelate inside the hydrogel. This interactions shorten the T1 and T2 relaxation times, inducing the switch-on of the contrast agent. This nanoswitch behaves as T1 contrast agent (r2/r1 = 1.8) at low field, whereas at high magnetic field it behaves as a T2 contrast agent (r2/ r1 = 21). Discussion/Conclusion: A highly efficient pH-nanoswitch for MRI has been synthesized based on the complexation between Mn cations and a pH-responsive nanogel. The rapid response to pH changes and the drastic decrease of T1/T2 offer great potential as contrast agent for in vivo pH imaging. The pH at which the swelling process is triggered, however, needs to be higher to be useful for in vivo

Purpose/Introduction: Double diffusion encoding (DDE) is a variant of diffusion-weighting in an MRI sequence.1,2 It contains a number of experimental parameters, whose influence on the MR signal is linked to tissue microstructure (see reference3). For both diffusion encodings, one can choose the gradient pulse duration (d), separation (D), gradient amplitude (G) and direction, as well as the [mixing-] time between encodings (sm). Varying these parameters individually is a way to estimate microscopic pore size4,5, or even the complete pore size distribution (PSD)6. Further experiments indicate that varying a combination of DDE parameters conjointly provides more stable and accurate PSD estimations7-9. This work aims at improving PSD estimation at clinically available gradient strengths by including a variable sm in such a multi-parametric protocol. Here, preliminary numerical simulation results are presented. Subjects and Methods: The expected signal attenuation for various PSDs was derived as a weighted sum of signal attenuations calculated for single-diameter pores. The whole dataset for each PSD is generated by repeating this process for all parameter sets under study. All numerical simulations were carried out with the MISST10-12 package in Matlab2016b (The MathWorks, Massachusetts, United States) using a spherical pore model. Table 1 shows the complete ranges of parameters simulated. Table 1 - MISST simulation parameters Parameter Min Increment Diameter (d) 5 μm 0.5 μm τm δ 1 ms δ 4 ms 1 ms Δ δ 1 ms Angle between encodings 0 rad π/8 rad Diffusion coefficient (D0) 2 μm2ms-1 G 45 mTm-1

Max 20 μm δ + 50 ms 24 ms δ + 50 ms 2π rad

PSD estimation was done via Tikhonov regularized (k) least-squares optimization and a non-negativity constraint: minimizex jjAx  yjj22 þkjjxjj22 ; such that xi [ 0 where x is the estimated PSD vector, y the [here simulated] measurement, and A the model matrix, constituted of the simulated

123

S634 single-diameter attenuation. Various signal-to-noise ratios (SNR) and parameter choices were tested. Results: Figure 1 shows two representative PSD estimations. The green line represents the ground truth, while the red bars show the result of the inversion. The accuracy of the inversion depends on the choice of PSD, acquisition parameters, SNR and k.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 diffusion NMR. J Magn Reson. 210 151–157 (2011). doi: 10.1016/j.jmr.2011.02.022. 11. Drobnjak I, Siow B, Alexander DC. Optimizing gradient waveforms for microstructure sensitivity in diffusion-weighted MR. J Magn Reson. 206 41–51 (2010). doi:10.1016/j.jmr.2010.05.017. 12. Ianus¸ A, Siow B, Drobnjak I, Zhang H, Alexander DC. Gaussian phase distribution approximations for oscillating gradient spin echo diffusion MRI. J Magn Reson. 227 25–34 (2013). doi: 10.1016/j.jmr.2012.11.021.

675 1 H, 13C and 19F hyperpolarization of three substituted pyridine derivates M. Plaumann1, R. Ringleb1, F. Euchner1, S. Hadjiali2, J. Bargon3, U. Bommerich1, G. Buntkowsky2, J. Bernarding1 1 Department for Biometrics and Medical Informatics, Otto-vonGuericke University, Magdeburg/GERMANY, 2Eduard-Zintl-Institute of Inorganic and Physical Chemistry, TU Darmstadt, Darmstadt/ GERMANY, 3Institute of Physical and Theoretical Chemistry, University of Bonn, Bonn/GERMANY

Discussion/Conclusion: From Figure 1, it can be seen that the inversion process is sufficient to discriminate between some PSDs. The SNR requirements being low, clinical applications might be feasible. Nonetheless, an optimal parameter choice (if it exists) is yet to be determined and implemented experimentally. Phantoms with specific PSDs also need to be created. References: 1. Cory DG, Garroway AN, Miller JB. Applications of Spin Transport as a Probe of Local Geometry. Polym Prepr. 31 149–150 (1990). 2. Mitra PP. Multiple wave-vector extensions of the NMR pulsedfield-gradient spin-echo diffusion measurement. Phys Rev B. 51 15074–15078 (1995). doi:10.1103/PhysRevB.51.15074. 3. Shemesh N, Jespersen SN, Alexander DC, et al. Conventions and nomenclature for double diffusion encoding NMR and MRI. Magn Reson Med. 75 82–87 (2016). doi:10.1002/mrm.25901. ¨ zarslan E, Basser PJ. MR diffusion-‘‘diffraction’’ phenomenon in 4. O multi-pulse-field-gradient experiments. J Magn Reson. 188 285–294 (2007). doi:10.1016/j.jmr.2007.08.002. 5. Koch MA, Finsterbusch J. Numerical simulation of double-wave vector experiments investigating diffusion in randomly oriented ellipsoidal pores. Magn Reson Med. 62 247–254 (2009). doi: 10.1002/mrm.21976. 6. Methot V, Koch MA. Pore size distribution estimation using the mixing time dependency of a double diffusion encoding experiment: a proof of concept from Monte Carlo simulated data. In: 33rd annual meeting ESMRMB, Vienna, Austria, 173 (2016). 7. Benjamini D, Komlosh ME, Holtzclaw LA, Nevo U, Basser PJ. White matter microstructure from nonparametric axon diameter distribution mapping. NeuroImage 135 333–344 (2016). doi: 10.1016/j.neuroimage.2016.04.052. 8. Katz Y, Nevo U. Quantification of pore size distribution using diffusion NMR: experimental design and physical insights. J Chem Phys.140 164201 (2014). doi:10.1063/1.4871193. 9. Katz Y, Benjamini D, Basser PJ, Nevo U. Reconstruction of Size Distribution of Cellular-Sized Pores Using DWI with Clinically Applicable Gradients. In: Proc. 23rd Annual Meeting ISMRM, Toronto, Canada. 2771 (2015). 10. Drobnjak I, Zhang H, Hall MG, Alexander DC. The matrix formalism for generalised gradients with time-varying orientation in

123

Purpose/Introduction: The hyperpolarizaton of pyridine was firstly presented by Adams et al. in 2009.1 Several derivatives were examined using the SABRE technique, since then. Currently, the first SABRE-based hyperpolarization studies in aqueous solution were published. Here, we present a detailed study in which the 1H, 13C and if feasible 19F hyperpolarization of 3-fluoropyridine, 4-methylpyridine and 3-fluoro-4-methylpyridine is compared. Subjects and Methods: The respective pyridine derivative (3-fluoropyridine, 4-methylpyridine and 3-fluoro-4-methylpyridine, Figure 1) was dissolved in 2 ml solvent in a 10 mm NMR tube. After addition of the Ir-catalyst (a) IMes-catalyst, b) Crabtree’s catalyst), the sample was degassed by using argon and and ultrasonic bath for 5 min. The hyperpolarization was realized with about 50% enriched parahydrogen and 6 bar pressure. Directly after hydrogenation at about 6mT, the NMR spectra were detected by using a single pulse experiment with a 90 excition pulse on a Bruker wide bore 300 MHz spectrometer. The obtained signal enhancements (SEs) were calculated from signal-to-noise ratios of the thermal and the hyperpolarized spectra.

Results: Strong signal enhancements are detectable for all examined substrates in the 1H NMR spectra. For example, the hyperpolarized 1 H and 19F NMR spectra (blue) as well as the spectra in thermal equilibrium of 3-fluoro-4-methylpyridine are shown below. When using the field cycling method, 13C signal enhancements could be detected. A comparison of the used catalyst systems demonstrates on the one side that best signal enhancements are observable when using the Ir-IMes catalyst. On the other side, changes of the phase of the increased signals are observable. Enhancements of more than 150 were calculated for the 1H signals.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S635

Novel Hardware and Sequences 676 8-Channel receive-only coil array for MR microscopy at 7 T E. Hosseini1, R. Frass-Kriegl2, L.I. Navarro De Lara2, J. Sieg2, M. Pichler2, E. Moser2, A. Berg2, E. Laistler2 1 Division MR Physics, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Wien/AUSTRIA, 2 Division MR Physics, Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Wien/AUSTRIA Purpose/Introduction: High resolution MRI strongly depends on the availability of highly sensitive radio frequency coils. The goal of this work is the development of a receiver coil array with high sensitivity and parallel imaging performance, that can be used in combination with an MR microscopy gradient insert for a 7T whole body scanner [1] and an available commercial birdcage transmit coil (Rapid Biomedical, Rimpar, Germany). Subjects and Methods: We designed and constructed a tube-shaped eight-channel receive-only array. The inner and outer diameter of the coil tube is limited by the space available in the transmit coil and the desired sample diameter, respectively (Fig. 1).

Discussion/Conclusion: Less is known about the hyperpolarization of 19 F. Until now, only the polarization of fluoro pyridine (using SABRE) is published.1,2 In the current study, we compare three pyridine derivates and within the influence of different functional groups concerning their effects on signal enhancements. Furthermore, a comparison of the catalyst system (Ir-IMes catalyst vs. Crabtree’s catalyst) shows interesting information like H/D exchanges and interaction with the solvent. The results give new important information for future studies in the field of hyperpolarization using the SABRE technique. References: 1. Adams RW, et al. Reversible Interactions with para-Hydrogen Enhance NMR Sensitivity by Polarization Transfer. Science, 2009; 323: 1708–1711. 2. Pravdivtsev AN, et al. Transfer of SABRE-derived hyperpolarization to spin-1/2 heteronuclei. RSC Adv., 2015; 5: 63615–63623.

The individual coil elements were arranged to cover the surface of the inner cylinder in two rows of four elements each. Elements within one row were mutually decoupled by transformer decoupling [2]. Decoupling between rows was achieved by coil overlap. In addition, preamplifier decoupling [3] was implemented. A double-layer flexible printed circuit board (PCB) was designed for each of the two rows. The coil conductors were deposited on one layer, the other layer contained solder pads for the circuitry required for tuning, matching, transformer decoupling, active detuning, and preamplifier decoupling. These PCBs were wrapped around the inner cylinder (phenolic paper) before soldering the components and could be moved with respect to each other to optimize overlap decoupling between rows. Preamplifiers are placed directly at the coil to minimize losses. Details of the coil design are shown in Fig. 2. The coil was tested on the bench.

123

S636

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

677 A quadrature surface loop receive/volume transmit coil design for rat brain MRI at 9.4 T and 14.1 T ¨ . Ipek1, P. Laub2, J. Fernandez Giacomini2, Y. Pilloud1, A. O Capozzi2, H. Lei1, R. Gruetter2 1 CIBM, EPFL, Lausanne/SWITZERLAND, 2Laboratory of Functional and Metabolic Imaging (LIFMET), Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Lausanne/SWITZERLAND

Results: Each element was successfully tuned to 297.2 MHz and matched to 75 X (noise match impedance of the used preamplifier). Efficient mutual decoupling (\-14 dB) was achieved by transformer and overlap decoupling (Fig. 3) and supplemented by preamplifier decoupling (\-15 dB). Active detuning circuits provided an isolation [40 dB between tuned and detuned state. The quality factor ratio of an unloaded and loaded array element was QU/QL = 1.6 for a cylindrical saline phantom (32 mm diameter, 0.6 S/m conductivity).

Discussion/Conclusion: An eight-channel receive-only coil, that can be used in combination with an MR microscopy gradient insert for a 7T whole body scanner and an available commercial birdcage transmit coil, was developed and tested on the bench. The use of thin flexible PCBs enables the incorporation of the coil circuitry and on-board preamplifiers despite the limited available space. As a next step, the coil performance will be evaluated in MR imaging experiments. References: [1] A. Berg, A. Potthast, P. Starewicz, MR-Microscopy on a human 7T-Scanner, Proc. ISMRM/ESMRMB (2010), p. 1048, Stockholm, Sweden. [2] N.I. Avdievich, H.P. Hetherington, 4 T Actively-Detuneable Double-Tuned 1H/31P Head Volume Coil and Four-Channel 31P Phased Array for Human Brain Spectroscopy, J. Magn. Reson. 186 (2007) 341–346. [3] A. Reykowski, S.M. Wright, J.R. Porter, Design of matching networks for low noise preamplifiers, Magn. Reson. Med. 33 (1995) 848–852. Acknowledgements: Austrian/French FWF/ANR Programme Blanc grant, Nr. I1371-B24, ‘‘FLEXAR7’’ Mayor of Vienna, Hochschuljubila¨umsfonds (HSJ) der Stadt Wien H-1208/2003, ‘‘MR-based dosimetry of High-LET radiation’’ (3D) proton and neutron dosimetry‘‘.

123

Purpose/Introduction: While surface coils enhances the transmit efficiency and receive sensitivity in the vicinity of the coils, a volume coil provides homogeneous excitation field over the large field-ofview1. By combining the volume and surface coils as an independent transmitter/receiver, an increased sensitivity close to the surface coils in a bigger field-of-view is expected2. This study aimed to design a volume transmit and quadrature surface receive coils for rat head imaging at 9.4 T (400 MHz) and 14.1 T (600 MHz) horizontal-bore MR scanners. Subjects and Methods: The coil structure is composed of two surface receiver loops (diameter = 2 cm each) and one copper Alderman-Grant coil (AG-coil). While the AG-coil diameter is kept as 5.5 cm, the length of the AG-coil is 3.8 cm for 9.4 T and 5.5 cm for 14.1 T. The tuning/matching and coupling parameters were measured on a network analyzer with a saline phantom tube (diameter = 2.6 cm, length = 11.5 cm) at 400 and 600 MHz. Active transmit/receive coil detuning was achieved with biasing and activation of PIN diodes on the coil circuitries. All phantom and in vivo experiments were carried out on a 14.1 T/26 cm and a 9.4 T/31 cm actively-shielded magnet (Varian/Magnex) with a saline phantom/ anesthetized adult rat (Sprague–Dawley, 253 g). The flip angle maps were acquired by using double-angle method3 (TR/TE = 15 s/ 2.77 ms, FA = 30, 60) and SNR maps acquired with GEMs sequence (matrix: 128 9 128, FOV: 45 9 45 mm, TR/TE = 200/ 13.8 ms, average = 1, Flip angle: 60 for signal and 0 for noise).

Results: The volume and surface coils were well-decoupled (better than-11 dB) by pin diode biasing while the quadrature surface coils were decoupled by overlapping (better than-11 dB). The quality factor of the loops were measured as 5.1 (unloaded/loaded = 112/22). Only at 14.1 T, the decoupling between one loop and AG-coil was not achieved (-10 dB) as it is visible as asymmetric GRE-images on the phantom. SNR is doubled for the quadrature surface receivers/volume transmit compared to the volume transmit/receive at 9.4 and 14.1 T.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Table1: Reflection and coupling parameters measured on the network analyzer for independent quadrature surface and volume transmit coil at 9.4T(400MHz) and 14.1T(600MHz). S11[dB],loop1 S22[dB],loop2 S33[dB],volume S12[dB],coupling S13[dB],coupling S23[dB],coupling [email protected] -2 -3 -24 -36 -20 -21 [email protected] -35 -39 -5 -13 -28 -14 [email protected] -5 -10 -26 -22 -28 -23 [email protected] -22 -26 -4 -11 -11 -19

S637

678 Space Harmonics Suppression (SHS) Wire Antenna: new RF probes for the Magnetic Resonance Imaging P. De Pellegars1, C. Coillot2, E. Nativel3, E. Alibert2, C. Goze-Bac2, R. Schimpf4, J. Muller5 1 Platforme BioNanoNMRI, SATT AxLR, Montpellier/FRANCE, 2 Laboratoire Charles Coulomb Plateforme BioNanoNMRI, University of Montpellier, Montpellier/FRANCE, 3UMR5214, Institut d’Electronique et des Syste`mes (IES), Montpellier/FRANCE, 4 President, RS2D, Mundolsheim/FRANCE, 5Chief Technical Officer, RS2D, Mundolsheim/FRANCE Purpose/Introduction: The use of Magnetic Resonance Imaging (MRI) in medical diagnosis and in the research community continues to grow in popularity because it is a non-invasive method. The imaging resolution is intimately linked to the probe efficiency producing the B1[1]. Moreover to optimize MRI studies, it is relevant to use MRI probe customized (Signal to Noise Ratio, homogeneity, sensibility…) to fit the sample shape[2]. Subjects and Methods: Traditionally RF probes are made up of inductive coils (solenoid, saddle coil, Helmholtz, birdcage…). However, some arrays of monopole or dipole wire antenna (resp. quarter and half wavelength) usually used with far field have been used on a high field MRI[3][9]. The monopole or dipole antenna of the array are evenly distributed around a cylinder[4]. Each antenna is associated to its own reception chain. SHS antenna consists in dipole antennas with several rungs/legs (by analogy with the coil turns of inductive probes). The optimal location h of the rungs is obtained by the space harmonics suppression of the current density (SHS)[7] to ensure good homogeneity of the magnetic field. The SHS antenna can be declined in two configurations. The asymmetric SHS antenna (Fig 1A) keeps the x–y plane homogeneity through the use of dipole antenna taking advantage of its natural 180 phase shift between the two parts put face to face. Homogeneity is improved with the number of rungs while the SNR is expected to be increased. Moreover, this antenna exhibits a linear gradient witch could act as MRI slice selection coil[8]. This is attributed to the current distribution due to propagative phenomenon in short leg[5]: I(z) = I0(1 - |z|/(L/2)). The symmetric configuration (Fig 1B), where the two SHS antennas fit into each other, allowing to compensate propagative phenomenon which results in a new type of promising homogeneous resonator.

Discussion/Conclusion: The quadrature loop receive/AG-volume transmit coils design shows increased FOV with a doubled SNR in the in vivo rat brain MR imaging at 9.4 T (Fig 3a). Despite decoupling challanges of the RF coils at 14.1 T, an increased receive sensitivity with an asymmetric pattern results in doubled SNR (Fig. 3b). References: (1) Collins, C. M., & Webb, A. G. (2007). Quadrature Surface Coils. In eMagRes. John Wiley & Sons, Ltd. Retrieved from http://onlinelibrary.wiley.com/doi/10.1002/ 9780470034590.emrstm1115/abstract. (2) Doty, F. D., Entzminger, G., Kulkarni, J., Pamarthy, K., & Staab, J. P. (2007). Radio frequency coil technology for small-animal MRI. NMR in Biomedicine, 20(3), 304–325. http://doi.org/10.1002/nbm.1149. (3) Cunningham, C. H., Pauly, J. M., & Nayak, K. S. (2006). Saturated double-angle method for rapid B1+ mapping. Magnetic Resonance in Medicine, 55(6), 1326–1333. http://doi.org/10.1002/mrm.20896.

Results: We have succesfully tested the 8 rungs asymmetric SHS on MRI 9.4 T (f0 = 400 MHz). Thanks to acquisitions on a sample of water we obtained a good linear gradient (Fig 2) and homogeneity in x–y plane (Fig 3).

123

S638

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [3] Raaijmakers AJ et al.: Design of a Radiative Surface Coil Array Element at 7 T: The Single-Side Adapted Dipole Antenna, Magn. Res. Med., 66(5), pp. 1488–1497, (2011). [4] Bolinger L., Prammer M. G. and Leigh J. S., A Multiple-Frequency Coil with a Highly Uniform B1 Field, Journal of Magnetic Resonance, Vol. 81, pp. 162–166, (1988). [5] « Antenna Fundamentals » lessons of National Radio Astronomy Observatory: http://www.cv.nrao.edu/course/astr534/AntennaTheory.html. [6] Hayes C. E., Edeutein W. A., Schenck J.F., Mueller O. M. and Eash M., An Efficient Highly Homogeneous Radiofrequency Coil for Whole-Body NMR Imaging at 1.5T, Journal of Magnetic Reso-nance, Vol. 63, pp. 622–628, (1985). [7] Coillot C., Nativel E., Zanca M. and Goze-Bac C., The magnetic fifield homogeneity of coils by means of the space harmonics suppression of the current density distribution, Journal of Sensors and Sensor Systems, 5(2), pp. 401–408, (2016). [8] John A.M. Lyon, Alan G-T Chaet Mohamed A. Hidayet, Tunable electrically small Antennas, Report of the Radiation Laboratory of Michigan University, (1971). [9] Hong, S.-M., Park, J. H., Woo, M.-K., Kim, Y.-B. and Cho, Z.-H. (2014), New design concept of monopole antenna array for UHF 7T MRI. Magn. Reson. Med., 71: 1944–1952. doi:10.1002/mrm.24844.

679 Dual channel power module for MRI shimming system J.P. Rigla Pe´rez1, G. P.casa´ns2, D. Grau-Ruiz2, L. Hernandez2, E. Diaz-Caballero1, J.M. Gonzalez2, H. Sa´nchez2 1 Research and development, Tesoro Imaging S.L., Alicante/SPAIN, 2 Instituto de Instrumentacio´n para Imagen Molecular (I3 M), Universitat Polite`cnica de Vale`ncia, Valencia/SPAIN

Discussion/Conclusion: There are various applications and uses for this new RF probe: *Agronomic or biologic applications using the high homogeneity of the asymmetric wire antenna. *Allows the possibility to switch between gradient or z-direction homogeneity during one acquisition (use the good possibility for the transmission and the reception depending on the study). *Could act as MRI slice selection coil. References: [1] Hoult D.I., Richards R.E., The signal-to-noise ratio of the nuclear magnetic resonance experiment, J Magn Reson, Vol. 24, pp. 71–85, (1976). [2] Meme, S., Joudiou, N., and Szmereta, F.: In vivo magnetic resonance microscopy of Drosophilae at 9.4T, Magn. Reson. Imag-ing, 31, 109–119, (2013).

123

Purpose/Introduction: Magnetic field homogeneity of the B0 field is crucial for good image acquisition in MRI. This is accomplished through the technic of active or passive shimming. We developed a low-cost multichannel current control power system (CCPS) for an active shimming system. The current control-loop (CCL) will be defined digitally rather than analogously [1, 2], as its usually done, to eliminate inaccuracies due to variations in the analog components. Subjects and Methods: The hardware developed consisted on a dual 20-port mixed signal driver MAX11301 for current control of 8 electromagnets through OPA549 operational amplifiers (Fig. 1a). Current flow is measured with a 0.1X resistor and the instrumentation amplifier AD622 with G = 10 V/V. Because the driver is I2C controlled, up to eight 8-ch systems can be interconnected for higher number of current sources. Control of the system is performed from a PC with the MCP2221 USB-I2C adapter. Flexibility on the current control loop is available as its control side is defined from the PC side rather than on the analog circuit (Fig. 1b). The design also considers the scenarios in which a current flow in the reverse direction is desired without altering the hardware connections. This is accomplished because of the symmetrical power supply consideration it was taken for such possible scenarios. Results: A test of the current control system was performed with an electromagnet coil. The control system was attached to the cold plate (set for 10 C) for proper refrigeration of the amplifier stage. Preliminary results show that the hardware developed can produce up to a 1.34 mT field with an increase of 11.7 C of the operational amplifier and of 10.7 C of the board for one channel being used (Fig. 1c). Discussion/Conclusion: Although there is flexibility in the current control-loop design, response time to changes is sacrificed. This can

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S639

be solved by using a microcontroller rather than a PC for system control definition and leaving the PC for parameter setup.

References: [1] N. Arango et al., Presented at 24th Annual Meeting ISMRM. [2] J.P. Stockmann et al., Magnetic Resonance in Medicine, vol. 75, no 1. pp 441–451, Jan. 2016.

680 Can scans with different TR be combined to improve UTE T2* measurements D.H.J. Poot, P. Baron, J. Hernandez Tamames Radiology and Nuclear Medicine Department, Erasmus MC, Rotterdam/NETHERLANDS

Experiments: The applicability is investigated by comparing with a simulation experiment evaluating S for T1 = 100–2000 ms, numerically finding the best matching FA2. Additionally, we perform an experimental validation by acquiring two images (FA1 = 8o, TR1 = 45.5 ms, FA2 = 5o, TR2 = 17.8 ms, TE = 0.032 ms, C = 0.6246, voxelsize = 0.586 9 0.586 9 2.00 mm3, FOV = 150 9 150 9 64 mm3) of a phantom consisting of 6 tubes differing in T1 (238–1312 ms) and T2 (17–49 ms)[2]. Results:

Purpose/Introduction: To enable combining UTE-SPGR sequences with different TR and increased number of short TE for T*2 quantification. UTE imaging is used to study tissues that have a compartment with short T*2(*1 ms) and longer T*2(*20–30 ms). To quantify both, images with short (\3 ms) and longer (*30 ms) TE have to be acquired. This requires multiple (multi-echo) UTE acquisitions, conventionally acquired with identical TR [ max TE. However, this scheme acquires unnecessarily many images with a long TE. Hence, we propose reducing the TR of some of the multi-echo scans, while preserving T1 weighting by appropriately choosing the flip angle (FA). Subjects and Methods: The UTE-SPGR signal is given in equation 1 (in figure 1)[1]. We want to use images with different TE, FA and TR to characterize f(TE), without having to quantify T1. Hence we want a relation between TR1, FA1, TR2, and FA2 such that the ratio of signals is independent of T1. Unfortunately, no closed form analytic relation obtaining that can be derived. Hence, the local relation with T1 (equation 2) is solved for small FA. A series expansion of that solution in FA1, TR2, and FA2, assuming small FA and TR  T1, is given by equation 3. Note that this is independent of T1. Figure 2 shows equation 3 and numerical solution. Up to FA1 = 39o, the approximation holds to good accuracy; \0.5o difference in FA. Using the numerical optimized values, the errors (equation 4) are less than 0.06% for all FA and T1.

123

S640

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 amplitude. Unlike previous methods [1],[3],[4], our constrained method involved no parameter tuning. This abstract expands upon this concept to design minimum out-of-slice error SMS pulses with direct peak, power, and in-slice error constraints. Subjects and Methods: We designed constrained RF pulses b that minimize out-of-slice error by using CVX [5] to optimize:

Figure 3 shows the ratio of the phantom images. It shows that this ratio is independent of the T1 and approximately 4% below the expected C. This small mismatch could be due to imperfect spoiling as the phantom does not have short T*2. As the difference is constant throughout the image, it can be compensated by calibration. Discussion/Conclusion: In this work we derived a relation between FA and TR that will allow combining UTE scans with different TR, without requiring knowledge on T1 or T2. This will enable, in the same scan time, increasing the number of short TE scans to better characterize the decay of short T*2 species. References: 1. M.A. Dieringer, M. Deimling, D. Santoro, J. Wuerfel, V.I. Madai, J. Sobesky, F. von Knobelsdorff-Brenkenhoff, J. Schulz-Menger, T. Niendorf, Rapid Parametric Mapping of the Longitudinal Relaxation Time T1 Using Two-Dimensional Variable Flip Angle Magnetic Resonance Imaging at 1.5 Tesla, 3 Tesla, and 7 Tesla, PLOS ONE, 2014, 9(3), e91318-e91318. 2. P. Baron, D.H.J. Poot, P.A. Wielopolski, E.H.G. Oei, J.A. Hernandez-Tamames, Accuracy of ADC measurements with an Ultrashort Echo Time Diffusion Weighted stimulated echo 3D Cones sequence, ISMRM 2017, program nr. 3452.

681 Minimum out-of-slice error SMS RF pules design with direct peak, power, and in-slice error constraints S.N. Williams1, J.A. Fessler2, D.C. Noll1 1 Biomedical Engineering, University of Michigan, Ann Arbor, MI/ UNITED STATES OF AMERICA, 2Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI/UNITED STATES OF AMERICA Purpose/Introduction: Simultaneous multislice (SMS) imaging accelerates MRI but is limited by peak RF amplitude and RF power deposition which scales roughly linearly with the number of simultaneous slices [1]. In [2], we showed a simple way for designing SMS small-tip angle (STA) RF pulses that directly constrained peak

123

Here, A is the STA approximation [6] system matrix, d is the target multislice excitation pattern, Win and Wout are the in-slice and out-ofslice regions, 4t is the dwell time, and bmax and pmax are the peak amplitude and power constraints, respectively. CSAR is a measured constant that converts integrated RF power to W/kg. The in-slice constraint emax is chosen to match the max absolute error of a comparable Shinnar Le-Roux (SLR) pulse. For the constrained design, we use the approach from [7] using a fractional transition width to define the transition band between Win and Wout. We compared our design to an SLR-based design (both least-squares (LS) and Parks-McClellan (PM) filters) [7] with optimized phase scheduling [1] and matched passband/stopband ripples. The design pulse length was chosen such that the SLR-based design met the peak amplitude constraint. We show simulation results for multiband 8, 0.3 mm, TBW = 4 pulses, with flip angle = 55, bmax = 0.2G, and pmax = 1.2 W/kg. Results: Figure 1 shows the three SMS RF pulses: our proposed constrained method, the SLR with LS filter, and the SLR with PM filter. Figure 2 shows the zoomed in magnetization magnitude profiles of a single slice for these methods. Table 1 compares the physical constraints, % max in-slice and % max out-of-slice errors for these designs.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S641

682 Detection of bone marrow changes related to osteoporosis using a stray field NMR U. Nevo Biomedical Engineering, Tel Aviv University, Tel Aviv/ISRAEL

Table 1. Peak RF amplitude, RF power, maximum % in-slice error, and maximum % out-of-slice error for SMS RF pulses. RF Power, Max % RF Peak Max % Error Error InRF Pulse Amplitude, CSAR||b||22△t Out-of-Slice Slice ||b||∞ (G) (W/kg) Constrained 0.2 1.2 3.7 1.1 Design SLR with LS 0.2 1.2 5.7 1.6 Filter SLR with PM 0.2 1.2 6.8 5.0 Filter

Discussion/Conclusion: Introducing direct constraints into the SMS design minimizes out-of-slice excitation error compared to conventional SLR-based SMS pulses while maintaining low in-slice error and meeting hardware limitations for peak RF amplitude and RF power. References: 1. Wong E. Optimized phase schedules for minimizing peak RF power in simultaneous multi-slice excitation pulses. Proc. Intl. Soc. Mag. Reson. Med. 2012; 20. 2. Williams SN, Noll DC, and Fessler JA. Improved Simultaneous Multislice Pulse Design Directly Constraining Peak RF Amplitude. Proc. Intl. Soc. Mag. Reson. Med. 2012; 25. 3. Aigner CS, Clason C, Rund A, Stollberger R. Efficient high-resolution RF pulse design applied to simultaneous multi-slice excitation. J. Mag. Reson. 2016; 263:33–44. 4. Sharma A, Lustig M, Grissom WA. Root-flipped multiband refocusing pulses. Mag. Reson. Med. 2016; 75(1):227–37. 5. Michael Grant and Stephen Boyd. CVX: Matlab software for disciplined convex programming, version 2.0 beta., September 2013. 6. Pauly J, Nishimura D, and Macovski A. A k-space analysis of small-tip-angle excitation. J. Mag. Reson. 1989; 81:43–56. 7. Pauly J, Le Roux P, Nishimura D, Macovski A. Parameter relations for the Shinnar-Le Roux selective excitation pulse design algorithm. IEEE Trans. Med. Imaging. 1991; 10(1):53–65.

Purpose/Introduction: Osteoporosis is characterized by reduction in trabecular bone in conjunction with increased marrow cell adiposity. While these changes occur within weeks, monitoring of treatment efficacy as performed by DEXA, is sensitive only to long-term changes. MRI is sensitive to bone-marrow changes, but is less affordable. Stray feild NMR scanners are a separate brancj of the NMR/MRI technology. These scanners are portable and use a low magnetic field twith permanenet magnets. Here we present our works on the use of a stray-field NMR device for the monitoring of an animal model of osteoporosis. Subjects and Methods: Two months-old female rats (n = 36) were ovariectomized (OVX) and their bone marrow compositiono was tracked using the NMR-MOUSE (Magritek, NZ), a stray field NMR device. Some of the rats were also dosed for the ensuing 3 or 5 weeks with 20 mg/kg of PTH(1–34). Hind limbs femurs and tibiae were isolated and underwent ex vivo micro-radiography and histology and NMR relaxometry at 6 weeks (preventive experiment) and 11 weeks (therapeutic treatment experiment) post OVX. NMR measurements were perfomed on ex vivo on the excised bones and included T1, T2 and diffusion measurements. Results: OVX rats showed significant changes in the values of T1, T2 and the diffusion, compared to SHAM rats. OVX rats developed osteoporotic changes including adipogenic marrow also compared to PTH treated rats. T2 and ADC NMR relaxation coefficients were found to correlate with marrow composition. Discussion/Conclusion: This study suggests that stray-field NMR, an affordable method that is sensitive to the rapid cellular changes in bone marrow, may have a clinical value in monitoring hormonal treatment for osteoporosis. References: Sarda Y*, Bergman E*, Hillel I*, Binderman I, Nevo U. Detection of Bone Marrow Associated with an Experimental Model of Osteoporosis Using a Stray Field NMR Scanner. In Press, Magn, Res. Med. 2016. Hillel I*, Sarda Y*, Binderman I, Nevo U. Monitoring of cellular changes in the bone marrow following PTH(1–34) treatment of OVX rats, using a portable stray field NMR scanner. In Press, J. of Osteoporosis, 2017.

123

S642

Peaks and Valleys - MR Spectroscopy 683 Effect of exercise on glycogen 13C-1 transverse relaxation time T*2 in human muscle at 7 T E. Sere´s Roig, R. Gruetter Laboratory of Functional and Metabolic Imaging (LIFMET), Ecole Polytechnique Fe´de´rale de Lausanne (EPFL), Lausanne/ SWITZERLAND Purpose/Introduction: In-vivo 13C-MRS provides unique insight into glycogen metabolism, such as the non-invasive detection of natural abundance glycogen-C1 in human muscle [1]. Moreover, the use of ultrahigh-field (i.e. 7 T) improves the 13C-sensitivity, while the application of 1H-decoupling remains a challenge due to power-limits. Advantageously for glycogen, the 1H-decoupling duration can be truncated accordingly to the fast 13C-T2 of glycogen-C1 [2], allowing its measurement at 7 T within power-limits [3]. Nonetheless, the 13CT2 of glycogen-C1 at 7 T may undergo changes during exercise, inasmuch as its 1H-T2 does it at 1.5 T [4]. Thus, the goal of this study was to investigate the exercise-dependence of the 13C-T2 of glycogenC1 in human muscle by 1H-decoupled 13C-MRS at 7 T. Subjects and Methods: All measurements were performed on a 7 T human-scanner (Siemens Erlangen/Germany) using a home-built 13Clinear/1H-quadrature surface coil [5]. A sphere (Ø = 7 mm) filled with 99% 13C-enriched formic-acid was placed in the centre of the 13 C coil as an external reference. A pulse-acquire sequence was developed for 1H-decoupled 13C-MRS with symmetric adiabatic 13Cexcitation [6] (2 ms) and broadband 1H-decoupling using the WALTZ16 scheme [7] (21 ms). Spectra were acquired from the calf muscle of a volunteer, pre- and post-exercise (1 h run), using the glycogen-C1 resonance at the centre of the spectrum. First and second orders FAST(EST) MAP [8] were applied pre- and post- exercise (VOI = 40 9 50 9 70 mm3). The 13C-T*2 of glycogen-C1 was calculated from the natural linewidth of the glycogen-C1 peak resonance, while all spectra were frequency and phase corrected. Results: Shimming pre- and post-exercise resulted in a water-linewidth of 26 Hz. Natural abundance of glycogen-C1 was detected at 100.5 ppm and its content was depleted by 40% post-exercise (Figure-1). The linewidths of two un-filtered spectra (not shown) acquired pre-exercise were 79 and 76 Hz, resulting in a mean 13C-T*2 of 4.1 ms for glycogen-C1 at rest. The linewidths of spectra acquired pre- and post-exercise were 125 and 98 Hz (Gaussian-filter = 50 Hz), respectively (Figure-1), resulting in a 22%-increase in 13C-T*2 postexercise.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion/Conclusion: The 13C-T*2 of glycogen-C1 in the rested muscle at 7 T (4.1 ms) is in agreement with the literature [9]. Because shimming was applied pre- and post-exercise, the post-exercise increment in 13C-T*2 due to macroscopic-susceptibility was discarded, while this was attributed to an increase in 13C-T2 rather than microscopic-susceptibility, the latter due the 1/2-spin of the 13C-isotope. In conclusion, the increased 13C-T2 remains within the 1H-decoupling duration permitted for SAR limits, conceding further investigation of this effect at 7 T. Plausible contributions to the exercise-induced 13CT2 increment may include alterations in tissue osmolality [4]. References: [1] Avison M.J.et al., PNAS-1988;85(5):1634–6. [2] Overloop K.et al., MRM-1996;36:45–51. [3] Sere´s Roig E.et al., 2015-ISMRM. [4] Price T.B. et al., J. Appl. Physiol. 1998;84(4):1178–84. [5] Sere´s Roig E. 2015-Thesis (EPFL). [6] Sere´s Roig E. et al., 2014 ISMRM/ESMRMB. [7] Shaka A.J. et al., JMR 1983; 53:313–340. [8] Gruetter R. MRM-1993;29,804–811. [9] Gruetter R. et al., MRM-1994;31:583–588.

684 Investigating TM-Averaged STEAM for Glutamate Measurements at 3 T R. Ka¨mpe1, A. Tisell2 1 CSAN, Linko¨ping University, IKE, Linko¨ping/SWEDEN, 2Radiation physics, Linko¨ping University, Linko¨ping/SWEDEN Purpose/Introduction: Glutamate (Glu), the main excitatory neurotransmitter of the central nervous system, has a complex spectrum with several strongly coupled and overlapping resonance (Glutamine (Gln)) [1]. Thus, Glu and Gln are virtually indistinguishable. However, separation is possible using TE averaged PRESS (TE-PRESS) [2]. But, the JPRESS signal will be smeared by the T2 relaxation between tTE potentially leading to reduction of quantification accuracy. The aim of this work was to investigate how the spin coupling affect the glu spectrum during the mixing time of a STEAM sequence and if it is possible to use mixing time (TM) averaged STEAM for separation of glu and gln. Subjects and Methods: Three sequences were investigated: TEPRESS, TE averaged STEAM (TE-STEAM) sequence with multiple tTE and TM averaged STEAM sequences with multiple tTM (TMSTEAM). Data was acquired on a 3 T Ingenia system (Philips Healthcare). Phantom measurements were performed using a Braino phantom (GE Healthcare, Chicago, USA) contain 12.5 mM Glu. Results: TE-PRESS (Fig. 1, first column) show a glu resonance and some editing effects at 2.2 ppm and a suppressed lactate resonance at 1.2 ppm. TE-STEAM (Fig. 1, second column) show similar spectra but less editing effects for both glu and lactate. TM-STEAM (Fig. 1, third column) show almost no editing effects, thus several glu resonances and a high lactate signal can be seen in the spectra. The spectra from TM-STEAM sequences with varying starting tTE values (Fig. 2) show that glu is not very affected by editing effects during the TM until a tTE of at least 80 ms is reached where some small editing effects can be seen at specific resonances. Lactate is almost completely suppressed by editing at a tTE of 144 ms (Fig. 3).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S643 STEAM sequence could potentially be used, just like a TE-PRESS sequence, to separate glutamate from glutamine and to avoid the smearing induced by the T2 relaxation between the different tTE values used in the TE-PRESS sequence. However, further validation on phantoms containing glutamate and glutamine is needed. References: [1] De Graaf, R. A. (2007). In vivo NMR spectroscopy: principles and techniques. Chichester, West Sussex, England; Hoboken, NJ, John Wiley & Sons. [2] Hurd, R., N. Sailasuta, R. Srinivasan, D. B. Vigneron, D. Pelletier and S. J. Nelson (2004). ‘‘Measurement of brain glutamate using TEaveraged PRESS at 3 T.’’ Magnetic Resonance in Medicine 51(3): 435–440. [3] Yang, S., J. Hu, Z. Kou and Y. Yang (2008). ‘‘Spectral simplification for resolved glutamate and glutamine measurement using a standard STEAM sequence with optimized timing parameters at 3, 4, 4.7, 7, and 9.4 T.’’ Magn Reson Med 59(2): 236–244.

685 Initial experience with a 3D MEGA-semi-LASER MRS sequence S. Tapper, A. Tisell, P. Lundberg Department of medical and health sciences, Linko¨ping University, Linko¨ping/SWEDEN

Discussion/Conclusion: We demonstrate that varying tTE has quite large editing effects on the glutamate for TE-PRESS. Just like Yang et al. [3] we also show that this editing effect is present when varying the tTM e.g. lactate signal is almost entirely suppressed and some glutamate resonances are affected. The lactate effect in combination with the small effect on glutamate (for tTE values higher than 80 ms with tTM steps of 1 ms, starting at 16 ms) indicates that a TM-

Purpose/Introduction: GABA is the main inhibitory neurotransmitter in the healthy human brain, and there are many contributions that have reported regional altered GABA concentrations in patients with neurological disorders. Moreover, it is highly advantageous to have access to a both reliable and non-invasive method when determining metabolite concentrations quantitatively for the study of neurological disorders. Therefore, a MRSI sequence intended for GABA quantification has been developed in collaboration with GyroTools (Zu¨rich, Switzerland). This sequence uses MEGA-semiLASER pulses for full brain coverage, minimal chemical shift displacement error, and uses spiral readout that minimizes the acquisition time. This sequence is still in the evaluation phase and as a result of the low SNR of GABA signal, only non-edited MRS is currently evaluated. Subjects and Methods: A 3T Philips Ingenia MR scanner and a standard GE spectroscopy phantom (Braino) have been used for data collections. Two-dimensional data (8 NSA, TR/TE = 1600/47 ms, water suppression VAPOR) were collected using a VOI (10 9 10 matrix, 10 9 10 mm2 in plane resolution, 10 mm slice thickness) placed in the center of the phantom. Directly afterwards, a shorter (2 NSA) unsuppressed water reference was collected using identical parameters and VOI placement. The data was reconstructed using ReconFrame (GyroTools, Switzerland) and the signals from the different coil elements were combined using SNR-weighting of the signals from the unsuppressed water reference. The data was also phase corrected according to Klose (1) and frequency corrected based on the residual water resonance. Afterwards, each spectrum was fitted using a Lorentzian function and concentration estimates were based on the area under the curve. From these concentration estimates, concentration maps were computed for the N-acetyl aspartate, Creatine and Choline signals.

123

S644 Results:

Fig. 1 illustrates the concentration maps computed from the measurement. Discussion/Conclusion: There are currently many challenges when using this sequence. First, the limitation of the B1 field, which may result in that the adiabatic pulses will not generate proper refocusing. Second, to achieve a short enough TE, the gradient spoiling will suffer and generate spurious echoes. Third, the spiral sampling requires offline reconstruction, which in 3D measurements in combination with the usage of a phased array coil generates huge data files. As described above, this MEGA-semi-LASER sequence is still in the initial evaluation phase, and there are several improvements that needs to be performed in the near future. However, when solving these issues, this sequence will be highly useful in clinical applications where small voxels and short acquisition times are desired. References: 1. Klose et al. 1990.

686 Exacerbated in vivo metabolic changes suggestive of a spontaneous muscular vaso-occlusive crisis in exercising muscle: a 31P MRS study in a mouse model of sickle cell disease

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 kinetics (p \ 0.05) during the post-stimulation recovery of one of the protocols. No significant difference was identified concerning the other protocol as well as muscle T2 values and volume. Discussion/Conclusion: Overall, our results showed significant alterations regarding phosphocreatine resynthesis during the postexercise recovery period in M1 as compared to the other SCD mice, which are suggestive of an acute ischemia. This study described, for the first time in a sickle cell mouse in vivo, exacerbated metabolic changes triggered by an exercise session that would be suggestive of an on-line observation of a muscular VOC although we did not provide evidence of a direct cause-effect relationship. References: [1] D.C. Rees, T.N. Williams, M.T. Gladwin, Sickle-cell disease. Lancet 376 (2010) 2018–31. [2] H.R. Schumacher, Jr., W.M. Murray, M.K. Dalinka, Acute muscle injury complicating sickle cell crisis. Semin Arthritis Rheum 19 (1990) 243–7. [3] V.C. Ejindu, A.L. Hine, M. Mashayekhi, P.J. Shorvon, R.R. Misra, Musculoskeletal manifestations of sickle cell disease. Radiographics 27 (2007) 1005–21. [4] B. Malekgoudarzi, S. Feffer Myonecrosis in Sickle Cell Anemia. New England Journal of Medicine 340 (1999) 483–483. [5] P. Vicari, R. Achkar, K.R. Oliveira, et al., Myonecrosis in sickle cell anemia: case report and review of the literature. South Med J 97 (2004) 894–6. [6] B. Giannesini, C. Vilmen, Y. Le Fur, et al., A strictly noninvasive MR setup dedicated to longitudinal studies of mechanical performance, bioenergetics, anatomy, and muscle recruitment in contracting mouse skeletal muscle. Magn Reson Med 64 (2010) 262–70.

687 Optimization of Echo Times for TE-Averaged PRESS Spectral Editing Technique Using Monte Carlo Simulations

B. Chatel1, L. Messonnier2, D. Bendahan1 1 CRMBM, UMR CNRS 7339, Aix-Marseille Universite´, Marseille/ FRANCE, 2LIBM, Universite´ Savoie Mont Blanc, Chambe´ry/ FRANCE

G.H. Hatay, E. Ozturk Isik Institute of Biomedical Engineering, Bogazici University, Istanbul/ TURKEY

Purpose/Introduction: Sickle cell disease (SCD) is a genetic hemoglobinopathy related to frequent vaso-occlusive crisis (VOC). If this type of crisis has been acknowledged in many tissues (1), the potential occurrence of such an event in skeletal muscle has been rarely documented (2–5). The present study reported exacerbated in vivo metabolic changes suggestive of a spontaneous muscular VOC in exercising muscle of a sickle cell mouse. Subjects and Methods: Anesthetized animal was placed supine within a home-built cradle especially designed for a strictly noninvasive investigation of posterior hindlimb muscles function and energetics (6). Using 31-phosphorus magnetic resonance spectroscopy, phosphocreatine and inorganic phosphate concentrations and intramuscular pH were measured in posterior hindlimb muscles throughout two standardized protocols of rest—exercise—recovery performed on different days at different intensities. These protocols were performed in ten SCD mice. Magnetic resonance imaging was used to assess muscle T2 values and volume at rest. Among these mice, one single mouse (referred as M1) displayed a very uncommon metabolic profile upon muscle stimulation. A statistical analysis (based on confidence intervals) was used to determine whether M1 was significantly different from other mice. Results: Our results illustrated that M1 had a slower phosphocreatine resynthesis rate (p \ 0.05) and inorganic phosphate disappearance

Purpose/Introduction: Proton magnetic resonance spectroscopic imaging (1H-MRSI) provides a noninvasive investigation of brain metabolism. Spectral editing techniques, such as TE-averaged pointresolved spectroscopy (PRESS)[1], enables separate quantification of glutamate (Glu), glutamine (Gln) and myo-Inositol (mI) by averaging spectra that are acquired with several echo times with 2D J-resolved PRESS [2]. Although TE-averaged PRESS is a powerful technique to separate Glu and Gln, it requires a long scan time that is proportional to the number of TE’s. The aim of this study is to optimize the echo times for TE-averaged PRESS to reduce its scan time using Monte Carlo simulations. Subjects and Methods: The spectral peaks of healthy brain metabolites, which are N-Acetyl aspartate (NAA), choline (Cho), creatine (Cr), ethanoliamine (Eth), g-Aminobutyric acid (GABA), Glu, Gln, mI, glycerol (Glyc), lactate (Lac), phenylalanine (Phe), pyruvate (Pyr), adenosine triphosphate (ATP) and taurine (Tau), were simulated using proton chemical shift, J-coupling and concentration values based on Govindaraju et al.’s [3] and Kraiser et al.’s [4] studies. 2D J-resolved PRESS was implemented in General Approach to Magnetic resonance Mathematical Analysis (GAMMA)[5], and each metabolite data file was separately created (minTE = 35 ms, DTE = 2.5 ms or 5 ms, 64 steps, 5000 Hz sweepwidth, 2048 points). Monte Carlo simulations were performed in MATLAB to optimize the minimum number of TE’s, whose average would provide the

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 maximum Glu and minimum Gln intensities at around 2.35 ppm. Keeping the clinical scan time limitations in mind, random combinations of 2, 3, 4 or 5 total number of TE’s were tested. After creating the TE-averaged PRESS spectra of selected TE’s, spectral peak areas were quantified in MATLAB. Results: Figure 1 shows the simulated spectral peaks of all brain metabolites separately for 35 ms and best TE combinations. Monte Carlo simulation results show that spectral peak of Glu was clearly separated from Gln at around 2.35 ppm with TE-averaged PRESS using up to 5 TE combinations (Figure 2). While optimizing for Glu, a small Gln peak that had the least contribution from Glu was observed at 2.43 ppm. Although the highest Glu/Cr ratio was obtained from a combination of 2 TE’s, the standard deviation and the range of combination results were slightly higher than the other number of TE combinations (Table 1).

S645

Discussion/Conclusion: Monte Carlo simulations were successfully performed to optimize the echo times for TE-averaged PRESS for a clinically feasible scan time. In future studies, the effect of noise, apodization and phase errors will be investigated. Further improvements for accelerating 2D J-resolved PRESS technique will be assessed to increase the number of TE’s. References: [1] Hurd, R. et al., Measurement of Brain Glutamate Using TEAveraged PRESS at 3T. Magn Reson Med, 2004. 51(3): p. 435–40. [2] Ryner, L.N., Sorenson, J.A., and Thomas, M.A., Localized 2D J-resolved 1H MR spectroscopy: strong coupling effects in vitro and in vivo. Magn Reson Imaging, 1995. 13(6): p. 853–69. [3] Govindaraju, V., Young, K., and Maudsley, A.A., Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed, 2000. 13(3): p. 129–53. [4] Kaiser, L.G. et al., A detailed analysis of localized J-difference GABA editing: theoretical and experimental study at 4 T. NMR Biomed, 2008. 21: p. 22–32. [5] Smith S.A., Levante T.O., Meier B.H., and Ernst, R.R. Computer Simulations in Magnetic Resonance. An Object Oriented Programming Approach. J. Magn. Reson., 106a.

688 Liver Cholesterol is a biomarker of Niemann-pick diseases progression and liver damage and could be quantified with MRS A. Xavier1, F. Zacconi2, K. Fuenzalida3, S. Zanlungo4, M. Andia5 1 Biomedical Imaging Center, Pontificia Universidad Cato´lica de Chile, Santiago/CHILE, 2Faculty of Chemistry, Pontificia Universidad Cato´lica de Chile, Santiago/CHILE, 3Instituto de Nutricio´n y Tecnologı´a de los Alimentos (INTA), Santiago/CHILE, 4 Faculty of Medicine, Pontificia Universidad Cato´lica de Chile, Santiago/CHILE, 5Radiology Department, Pontificia Universidad Catolica de Chile, Santiago/CHILE Purpose/Introduction: Niemann-Pick (NP) is a genetic and rare group of diseases that affect many body organs. NP patients have an abnormal lipid metabolism that causes an accumulation of lipids in liver, spleen, brain and bone marrow. NP patients develop cirrhosis very fast [1]; therefore this disease would provide information about the mechanisms and biomarkers of fatty acid damage in the liver. The purpose of this study is to characterize the liver fatty acids composition using MRS and gas chromatography (GC) in a mice model of Niemann-Pick type C and compare with wild-type (WT) and non alcoholic fat liver disease (NAFLD) mice with the aim of provide a non-invasive technique to follow-up this patients and provide new hypothesis about the bad progression of NAFLD in some patients.

123

S646

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Subjects and Methods: Livers from NP type C mice were analyzed (N = 4). WT mice and NAFLD mice of the same age were analyzed as controls. Fatty Acids Methyl Esters (FAME) from the mice liver were extracted [2] and analyzed using a 9.4 T MRS (Bruker Ultrashield 400 MHz) and gas chromatography with a mass spectrometer (Agilent technologies). Results: Concentration measured in the gas chromatography of the palmitic (C16:0), stearic (C18:0), oleic (C18:1) and arachidonic (C20:4) acids methyl esters were significantly different between the NPC livers and the NAFLD and WT livers (Figure 1). Others FA detected did not have significant differences. Additionally, a great quantity of cholesterol was found in NPC livers and no cholesterol was found in the NAFLD and WT livers (Figure 2). In the MRS (Figure 3) the peaks in 0.87 and 0.97 ppm, which are related to total cholesterol and free cholesterol structure, showed a significant difference between NPC livers and NAFLD and WT livers. The peaks related with fatty acids structure, like the beta-carboxyl methylene (1.61 ppm) and alpha carboxyl methylene (2.28 ppm) also showed significant differences. The rest of the MRS fat peaks did not show significant differences.

Discussion/Conclusion: Our results showed that the increase in the liver cholesterol concentration is the main difference of NPC liver manifestation and it has been associated with the bad liver prognosis. Also, there is a different pool of fatty acids storage in the NPC liver, and the different MR liver spectra would provide biomarkers for noninvasive follow-up of these patients. The accumulation of cholesterol in the liver would be a biomarker of bad prognosis of NAFLD and could be detected not invasively using MRS. References: [1] BELTROY E et al. Hepatology 2005. [2] FOLCH, J. STANLEY, S. J. Biol. Chem. 1956. Acknowledgment: CONICYT-PIA, Anillo ACT1416.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

689 Old dog, new tricks—Functional, non-destructive and quantifying MRI/MRS methodology of bones and soft tissue in a rat model for degradable magnesium based implants M. Meier1, D. Haake1, A. Weinberger2, R. Willumeit-Ro¨mer3 1 ZTL-Imaging Center, Hannover Medical School, Hannover/ GERMANY, 2Musculo-Skelettale FE f. Biomaterialien, Medical University of Graz, Graz/AUSTRIA, 3Biologische Grenzfla¨chen von Implantaten, Christian-Albrechts-University Kiel, Kiel/GERMANY Purpose/Introduction: In orthopedics, an MRI may be used to examine bones, joints, and soft tissues such as cartilage, muscles, and tendons. This project developed MRI and MRS methods for determining the quality of the degradation process of implants in bone. Magnesium based implants are valuable because Mg is an essential element in the body, mechanical properties are similar to bone, it should enhance bone formation and it is a promising material for the growing skeleton of children. The fundamental understanding of degradation control in different environments is still missing. Subjects and Methods: We used a rat model for studying different magnesium based degradable implants and compared those with non degradable material and control treatment. We used diffusion imaging, perfusion imaging, T1- and T2-mapping and UTE-methods to correlate with histology. Spectroscopy was used to characterize the degradation process and the healing process after implantation. Results: This study demonstrated the superior nature of 3D MRI in early quantitation of fibrosis between the host bone and the implant. The presence of such fibrous membranes is indicative of failure of the implant. This study provided a measure of the distortions in the morphological bone parameters derived from MR images due to susceptibility artefacts and partial volume effects. MRS revealed signals of tCr (3.03 ppm), Trimethyl ammonium compounds TMA (3.21 ppm), Lipid resonances (0.9–1.4 ppm), a Smaller tCr peak, a transient Lactate signal and a short time signal of deoxymyoglobin = (de)oxygenation. The desire to study bone post-operatively led to determine the extent of tissue that could be investigated around the implant. Discussion/Conclusion: When using MRI to monitor implant degradation and surrounding tissue, the image quality is severely affected by the presence of the metals. This problem was also investigated by simulations of the effects of slice excitation and frequency encoding. The simulations could predict the effects of imaging parameters on artefact size. It could be shown that 2D SE imaging techniques should not be used with metallic implants due to extreme slice distortion, whereas 3D MRI provided a method that has effectively no slice distortion, although the effects of frequency encoding gradient still remain. The corrosion of magnesium is accompanied by hydrogen evolution and a local increase in pH, which impose constraints on many potential biomedical applications. We have overcome these limitations with the next generation of metallic biodegradable implant materials with the addition of biocompatible elements. Along with the addition of Ca, involved in bone formation and remodeling, excellent material properties were achieved. References: Witte, Kalla, Meier. European Cells and Materials Vol. 23. Suppl. 2, 2012. Xu, Yu, Zhang, Pan, Yang. Journal of Biomedical Materials Research DOI: 10.1002/jbm.a.31273. Zhang, Yifeng et al. Nature medicine 22.10 (2016): 1160–1169.

S647

689a TowardsOpto-fMRS: Assessment of metabolic changes resulting from optogenetic stimulation using a BOLD free-difference spectrum N. Just, C. Faber Mu¨nster, DE, University Hospital of Mu¨nster, Translational Research Imaging Center Purpose/Introduction: The assessment of metabolic changes resulting from brain activation using a BOLD-free difference spectrum is becoming a prerequisite in1H functional MR spectroscopy (fMRS) (1,2). The procedure is however difficult in rodents as high SNR and well shimmed spectra in small voxels with improved temporal resolution remain challenging. The implantation of an optical fiber following viral transduction is not supposed to ease this task for optofMRS. Here, a BOLD-free difference spectrum was obtained following green laser stimulation of the rat forepaw cortex. Subjects and Methods: An optical fiber (OF) was implanted on the day of fMRS into the S1FL of virally transduced Fisher rats (injection of AAV2-CamKII-C1V1 four weeks prior). Animals were intubated and placed in a stereotactic cradle. Experiments were conducted at 9.4 T (Biospec, Bruker) under medetomidine sedation. Body temperature (37 ± 1C) and respiration rate were continuously monitored. Optogenetic stimulations (1 ms, 9 Hz) were performed using a repeated 5 min OFF 5-min ON block paradigm. Green laser light was delivered through the OF. STEAM 1H-fMRS was conducted with optimized OVS and VAPOR pulses. FASTMAP-shimmed spectra were acquired in a 10 ll VOI localized over the activated S1FL region previously determined with BOLD fMRI and encompassing the tip of the OF. BOLD fMRI series were analyzed using SPM12. 1 H-MR spectra were processed using in-house written Matlab routines. Results: FIDs were summed over 5-min periods of optogenetic stimulation and rest respectively per animal (n = 5). Then, stimulated and resting 1H spectra were averaged across rats (Fig. 1). The averaged stimulated spectrum demonstrated increased NAA and tCr peak heights (+3.83% for NAA and +1.61% for tCr) reflecting BOLD responses. The BOLD effect was eliminated by 0.2 Hz linebroadening on NAA and tCr peaks of the stimulated spectrum in order to match the linewidth of the rest spectrum and a difference spectrum was obtained (Fig. 2). Negative peaks for glucose and aspartate and positive peaks for lactate, glutamate and potentially alanine were found. Discussion/Conclusion: We successfully obtained a BOLD-free difference spectrum upon optogenetic stimulation of the rat forepaw cortex. Preliminary findings are in line with previous findings in rat (2) but will need to be confirmed with a larger population size and quantification of metabolite changes. Reliable quantification of metabolite concentration is currently under progress. Early results warrant further development of this new tool to evaluate brain metabolism during activation.

123

S648

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 texture features were extracted from each ROI using PyRadiomics [2] with a constant bin width (=2), and relative minimum and maximum. Spearman’s correlation coefficient (r) was estimated between each feature computed in ROI_a and ROI_b and the corresponding volumes for the 30 subjects. Wilcoxon rank sum test followed by Benjamini–Hochberg correction was used to compare: 1) features computed in ROI_b and ROI_c, 2) features computed in ROI_c and ROI_d.

References: 1. Schaller B et al. Net increase of lactate and glutamate concentration in activated human visual cortex detected with magnetic resonance spectroscopy at 7 tesla. J Neurosci Res. 2013;91:1076–83. 2. Sonnay S, Duarte J, Just N. Lactate and glutamate dynamics during prolonged stimulation of the rat barrel cortex suggest adaptation of cerebral glucose and oxygen metabolism. Neuroscience. 2017; 27;346:337–348.

Rise of the Machines 690 Impact of volume size on radiomic features computed from MR images: an illustration in pediatric neuroimaging J. Goya Outi1, F. Orlhac1, R. Calmon2, C. Nioche1, C. Philippe3, A. Alentorn3, J. Grill4, V. Frouin3, F. Frouin5 1 IMIV, INSERM/CEA/CNRS/Universite´ Paris-Sud/Universite´ ParisSaclay, ORSAY/FRANCE, 2Radiologie, AP-HP, Paris/FRANCE, 3 Neurospin/DRF, CEA, Gif-sur-Yvette/FRANCE, 4CNRS UMR 8203, Gustave-Roussy, Villejuif/FRANCE, 5INSERM/CEA-SHFJ/CNRS/ Universite´ Paris-Sud/Universite´ Paris-Saclay, IMIV, ORSAY/ FRANCE Purpose/Introduction: The advent of radiomic studies generates a lot of enthusiasm and development in the medical imaging community. However, the real impact of this new field can be hampered by an inappropriate use of tools. This study illustrates the impact of the volume of regions of interest on which radiomic features are computed and proposes recommendations to avoid misinterpretation of texture indices. Subjects and Methods: Morphometric MR scans (T1, post-contrast T1, T2, and FLAIR) of thirty children (3–15 years old) with diffuse intrinsic pontine glioma were analyzed. Image intensities were standardized according to a specific procedure, inspired from [1] and images sampled on a cubic voxel grid (1 mm3) were used. Using the diffeomorphic transformation estimated to normalize each volume to the MNI template, two concentric spherical regions (3 and 5 mm radius) drawn in the white matter of the MNI-brain were transferred to subjects’ images (ROI_a, ROI_b, Figure 1). The resulting ROI volumes (39–424 mm3) varied between subjects according to their age and brain size. Furthermore, two spherical ROIs, ROI_c and ROI_d (4.5 mm radius), were drawn in the patients’ space: ROI_c, ROI_a and ROI_b were concentric and ROI_d mirrored ROI_c in the opposite hemisphere (Figure 1). For each of the four modalities, 60

123

Results: A total of 46/240 features presented high correlation with volume (|r| [ 0.7). Among the 60 features, 8 were systematically correlated with volume, whatever the MR modality. Figures 2 and 3 illustrate three features varying differently with the volumes. A total of 81 features presented significant differences (adjusted p-value \ 0.05) when computed in ROI_b and ROI_c, while no significant difference was found between features computed in ROI_c and ROI_d (minimal p-value 0.84).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S649

691 WITHDRAWN 692 3-D LGE-MRI Segmentation using a Random Forest Classifier and Dynamic Programming T. Kurzendorfer1, A. Brost2, C. Forman3, M. Schmidt3, C. Tillmanns4, S. Steidl1, A. Maier1 1 Pattern Recognition Lab, Friedrich-Alexander University ErlangenNuremberg, Erlangen/GERMANY, 2Advanced Therapies, Siemens Healthcare GmbH, Forchheim/GERMANY, 3Magnetic Resonance, Siemens Healthcare GmbH, Erlangen/GERMANY, 4KardioDiagnostik, Diagnostikum Berlin, Berlin/GERMANY Purpose/Introduction: Ischemic heart disease is the leading cause of death (1). State-of-the-art is 2-D LGE-MRI to visualize myocardial scar, however, incomplete coverage doesn’t allow precise scar quantification. Therefore, an isotropic high resolution 3-D whole heart LGE-MRI was proposed (2). The challenge lies in the delineation of the myocardium. We developed a method for automatic left ventricle (LV) segmentation in 3-D LGE-MRI based on random forest (RF) classification and dynamic programming as pre-requisite for scar quantification. Subjects and Methods: The segmentation of the LV consists of five steps. First, the LV is initialized using a two-stage registration (Figure 1B)(3). Second, the short axis view is estimated using principal component analysis (Figure 1C).

Third, the endocardium is estimated using a RF classifier, which is trained using 16 steerable features (4). Potential boundary points are extracted using circular ray casting (Figure 2A). The classification result is interpreted as boundary costs (Figure 2B, 3A).

Discussion/Conclusion: This study demonstrates the dependence of some texture indices on region volume. When ignoring this effect, statistical conclusions might be erroneous. Computing texture parameters from regions of similar shape and size and standardized intensities could enhance robust inter-patient comparisons. References: [1] Shinohara et al. Statistical normalization techniques for magnetic resonance imaging. Neuroimage Clinical 2014;6:9–19. [2] PyRadiomics, Python package v1.1.1, available at: (Feb 2017).

123

S650

To improve the endocardial segmentation, a scar exclusion is added. Therefore, the mean intensity and standard deviation of the blood pool are estimated. The scar threshold is defined as the standard deviation plus the mean intensity (Figure 3B). A scar map is generated, where all pixels with increasing radius from potential scar candidates are labeled with 1 (Figure 3C). The scar map is combined with the boundary map, resulting in a cost map (Figure 3D). The shortest path in polar space is found using a minimal cost path (MCP) (Figure 3E).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 The segmentation was evaluated using 9 LGE-MRI data sets [sparse GRE sequence and reconstruction, spatial resolution (1.3 mm)3]. A comparison to gold-standard annotations from two clinical experts using a leave-one-out cross-validation was performed. Results: The segmentation resulted in a Dice coefficient of 0.84 ± 0.07 for the endocardium and 0.85 ± 0.05 for the epicardium. The biggest differences occur at the left ventricular outflow tract. Incorporating a heart model might further improve the segmentation. Discussion/Conclusion: The presented work segments the LV solely using LGE-MR. It has been shown that a trained RF classifier for the endocardium and epicardium combined with a MCP can lead to good LV segmentation results, as a prerequisite for scar quantification. References: Disclaimer: The methods and information presented in this paper are based on research and are not commercially available. 1. Nichols, Melanie; Townsend, Nick; Scarborough, Peter; Rayner, Mike. Cardiovascular disease in Europe: epidemiological update. January 2013, Vol. 34, 39. 2. T. Shin; M. Lustig; D. Nishimura; B. Hu. Rapid single-breath-hold 3D late gadolinium enhancement cardiac MRI using a stack-of-spirals acquisition. Journal of Magnetic Resonance Imaging. 40, 2014, Vol. 6. 3. Tanja Kurzendorfer, Christoph Forman, Michaela Schmidt, Christoph Tillmanns, Andreas Maier and Alexander Brost. Fully Automatic Segmentation and Scar Quantification of the Left Ventricle in 3-D Late Gadolinium Enhanced MRI. 11th Interventional MRI Symposium. 2016. 4. Zheng, Yefeng; Barbu, Adrian; Georgescu, Bogdan; Scheuering, Michael; Comaniciu, Dorin. Four-chamber heart modeling and automatic segmentation for 3-D cardiac CT volumes using marginal space learning and steerable features. IEEE transactions on medical imaging. 27, 2008, Vol. 11. 5. Karim, Rashed; Bhagirath, Pranav; Claus, Piet; Housden, R James; Chen, Zhong; Karimaghaloo, Zahra; Sohn, Hyon-Mok; Rodrı´guez, Laura Lara; Vera, Sergio; Alba`, Xe`nia and others. Evaluation of stateof-the-art segmentation algorithms for left ventricle infarct from late Gadolinium enhancement MR images. 2016, Vol. 30.

693 Automatic segmentation of knee muscles in 3D MRI data using deep learning K. Giang1, A. Chodorowski2 1 Electrical Engineering, Signals and systems, Chalmers University of Technology, Gothenburg/SWEDEN, 2Electrical Engineering, Signals and Systems, Gothenburg/SWEDEN

The contour is transformed back to Cartesian coordinates and the convex hull is obtained (Figure 2C, D). These steps are repeated for each slice, until base and apex are reached. Afterwards, the epicardial boundary is estimated using the RF classifier and the MCP (Figure 2E). In the final step, the myocardium contour can be exported and used for scar segmentation.

123

Purpose/Introduction: An international survey among 18 countries (755 million people) showed that 1,324,000 patients undergo knee arthroplasty annually [1]. Higher accuracy of prosthesis placement in surgical outcome can be achieved through image guided surgery, which require image segmentation of anatomical structures around the knee, in particular the muscles. During recent years, convolutional neural networks (CNN) has excelled in image recognition and segmentation tasks, also within medical segmentations in MRI [2, 3]. Our aim was to create and evaluate an automatic segmentation algorithm based on deep learning for muscles around the knee in 3D MRI data. Subjects and Methods: The data was composed of 18 knee 3D fast spin echo, proton density weighted images with volume dimension 512 9 512 9 256, voxel spacing 0.3125 9 0.3125 9 0.6 mm (GE Medical Systems 3T Discovery MR750 scanner) [4, 5]. The 3D CNN was created in Python 3.6.0 using the libraries Lasagne v.0.2.dev1 and Theano v.0.9.0 with cuDNN 5110.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S651

In the algorithm the volumetric MRI data were sampled voxelwise, and a 3D patch of size 28 9 28 9 28 voxels was created around each sampled center voxel. The patches were used as input to the network trained by backpropagation guided by expert annotations. Output from the network was a probability of the voxel either belonging to foreground (muscle) or background (non-muscle). In order to reconstruct the volumetric image, each single voxel was sampled and forward passed into the network resulting in a likelihood map. Fig. 1 shows the CNN architecture used for this segmentation. A global thresholding on the likelihood map followed by morphological opening resulted in the final segmentation. The constructed CNN classifier was trained using 14 patients, two validation and two testing, and evaluated with Dice similarity coefficient.

Results: The resulting volumetric Dice coefficient on the test data was 0.88 for muscle voxels (foreground) and 0.97 for background voxels, in the whole volume. Mean slice per slice Dice coefficient can be seen in Fig. 3. It demonstrates that the 3D CNN has easier to classify non-central and non-end muscles. An example slice segmented is shown in Fig. 2, the brighter voxel, the higher probability of it being classified as muscle.

Discussion/Conclusion: The obtained level of segmentation accuracy suggests that 3D CNN based segmentation may be used in surgical planning and treatment. However, to avoid too optimistic results and overfitting, the classifier should be evaluated on a larger dataset and by k-folded cross-validation. Also the only postprocessing in the segmentation were through thresholding and morphological opening, which leaves room for improvement. References: [1] Kurtz M. S., Ong L. K., Lau E., et. al. (2011). ‘‘International survey of primary and revision total knee replacement’’. Springer, International Orthopaedics. pp.1783–1789. [2] B. V. Ginneken., ‘‘Deep learning approaches in MRI—and beyond…’’, ESMRMB Congress, Book of Abstracts, Vienna, 29, p.253, 2016. [3] K. Kamnitsas et al., ‘‘Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation’’, Medical Image Analysis 36, 61–78, 2017. [4] M. Lustig, Vasanawala S. http://mridata.org/fullysampled, accessed: 2016-07-07. [5] K. Epperson, et al., ‘‘Creation of Fully Sampled MR Data Repository for Compressed Sensing of the Knee,’’ SMRT Conference, Salt Lake City, UT, 2013.

694 3D Convolutional neural network for hippocampal subfiels segmentation in ultra-high resolution MRI N. Jacobsen1, B.D. Hansen1, A.K. Nøhr1, L.R. Østergaard1, S.B. Petersen1, S. Bollmann2 1 Department of health science and technology, Aalborg University, Aalborg/DENMARK, 2Centre for Advanced Imaging, The University of Queensland, St Lucia/AUSTRALIA Purpose/Introduction: Recent studies suggest that volumetric measurements of hippocampal subfields could deliver biomarkers for an early detection of Alzheimer’s Disease [1–6], and increased access to

123

S652 7 T MRI have made segmentation of these subfields feasible [7, 1]. Atlas-based methods for automatic segmentation of the hippocampal subfields have been proposed. However, these methods show limitations with respect to segmentation of smaller subfields and they are very time consuming [8–10]. In current research, Convolutional Neural Networks (CNN) have been used to segment brain structures and lesions, and the approach shows fast and accurate segmentations [11–18]. This study aims to explore CNNs as a fast and reliable method for hippocampal subfield segmentation. Subjects and Methods: We used MP2RAGE and TSE contrast minimum deformation average models [18–20] to generate highquality training labels. To achieve this we segmented the subfields in the high-resolution model space using FreeSurfer and transferred the labels to 3 healthy subjects utilizing the transformation matrices obtained during the average model construction. To further augment the dataset we flipped the 3 subjects and the corresponding labels. The proposed segmentation method is a supervised 3D CNN developed using the framework DeepMedic [11]. The network has two parallel pathways consisting of 8 convolutional layers, 2 fully-connected layers, a softmax layer, and a CRF layer. The network was trained using an approach based on dense inference [11]. The network was trained on 5 subjects and tested on the left-out subject (Leaveone-out cross validation) to achieve a segmentation performance measurement. The predicted labels were compared to the training labels using dice scores. Results: Figure 1 represents average dice score through a training period of 15 epochs for each of the 6 subtests. The total average dice score was 0.9031 ± 0.0182. Training time was between 19 and 20 h. Figure 2 illustrates the predicted segmentations for subtest 5 after 15 training epochs, compared to ground truth labels.

Discussion/Conclusion: It was possible to segment hippocampal

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 subfields with an average dice score of 0.9031 ± 0.0182. As illustrated in figure 2 the labels used as ground truth show some inconsistency in the boundaries for the subfields, reducing the requisite of the CNNs training and thereby the quality of the predicted segmentations. The high precision regardless of the ground truth labels shows that CNNs with the right configurations and regularization methods are highly adaptable to new tasks, and perform well on segmentation tasks despite small datasets. Increasing the training dataset could reduce errors further and increase accuracy. References: [1] C. Boutet, M. Chupin, S. Lehe´ricy, L. Marrakchi-Kacem, S. Epelbaum, C. Poupon, C. Wiggins, A. Vignaud, D. Hasboun, B. Defontaines, O. Hanon, B. Dubois, M. Sarazin, L. Hertz-Pannier, and O. Colliot, ‘‘Detection of volume loss in hippocampal layers in alzheimers disease using 7 T MRI—a feasibility study,’’ NeuroImage: Clinical, vol. 5, pp. 341–348, 2014. [Online]. Available: http://dx.doi.org/10.1016/j.nicl.2014.07.011. [2] G. A. Kerchner, C. P. Hess, K. E. Hammond-Rosenbluth, D. Xu, G. D. Rabinovici, D. A. C. Kelley, D. B. Vigneron, S. J. Nelson, and B. L. Miller, ‘‘Hippocampal CA1 apical neuropil atrophy in mild alzheimer disease visualized with 7T MRI,’’ Neurology, vol. 75, no. 15, pp. 1381–1387, oct 2010. [Online]. Available: http://dx.doi.org/10.1212/WNL.0b013e3181f736a1. [3] R. L. Joie, A. Perrotin, V. de La Sayette, S. Egret, L. Doeuvre, S. Belliard, F. Eustache, B. Desgranges, and G. Che´telat, ‘‘Hippocampal subfield volumetry in mild cognitive impairment alzheimer’s disease and semantic dementia,’’ NeuroImage: Clinical, vol. 3, pp. 155–162, 2013. [Online]. Available: http://dx.doi.org/10.1016/j.nicl.2013.08.007. [4] A. Maruszak and S. Thuret, ‘‘Why looking at the whole hippocampus is not enough—a critical role for anteroposterior axis, subfield and activation analyses to enhance predictive value of hippocampal changes for alzheimer’s disease diagnosis,’’ Frontiers in Cellular Neuroscience, vol. 8, mar 2014. [5] J. Pluta, P. Yushkevich, S. Das, and D. Wolk, ‘‘In vivo analysis of hippocampal subfield atrophy in mild cognitive impairment via semiautomatic segmentation of t2-weighted mri,’’ Journal of Alzheimer’s Disease, vol. 31, no. 1, pp. 85–99, 2012. [6] A. G. van der Kolk, J. Hendrikse, J. J. Zwanenburg, F. Visser, and P. R. Luijten, ‘‘Clinical applications of 7 T MRI in the brain,’’ European Journal of Radiology, vol. 82, no. 5, pp. 708–718, may 2013. [Online]. Available: http://dx.doi.org/10.1016/j.ejrad.2011.07.007. [7] L. E. M. Wisse, H. J. Kuijf, A. M. Honingh, H. Wang, J. B. Pluta, S. R. Das, D. A. Wolk, J. J. M. Zwanenburg, P. A. Yushkevich, and M. I. Geerlings, ‘‘Automated hippocampal subfield segmentation at 7T MRI,’’ American Journal of Neuroradiology, vol. 37, no. 6, pp. 1050–1057, feb 2016. [Online]. Available: https://doi.org/10.3174%2Fajnr.a4659. [8] J. E. Iglesias, J. C. Augustinack, K. Nguyen, C. M. Player, A. Player, M. Wright, N. Roy, M. P. Frosch, A. C. McKee, L. L. Wald, B. Fischl, and K. V. Leemput, ‘‘A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI,’’ NeuroImage, vol. 115, pp. 117–137, jul 2015. [Online]. Available: https://doi.org/10.1016%2Fj.neuroimage.2015.04.042. [9] A. Giuliano, G. Donatelli, M. Cosottini, M. Tosetti, A. Retico, and M. E. Fantacci, ‘‘Hippocampal subfields at ultra high field MRI: An overview of segmentation and measurement methods,’’ Hippocampus, vol. 27, no. 5, pp. 481–494, feb 2017. [Online]. Available: https://doi.org/10.1002%2Fhipo.22717. [10] H. Choi and K. H. Jin, ‘‘Fast and robust segmentation of the striatum using deep convolutional neural networks,’’ Journal of Neuroscience Methods, vol. 274, pp. 146–153, dec 2016. [Online]. Available:

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [11] K. Kamnitsas, C. Ledig, V. F. Newcombe, J. P. Simpson, A. D. Kane, D. K. Menon, D. Rueckert, and B. Glocker, ‘‘Efficient multiscale 3d cnn with fully connected crf for accurate brain lesion segmentation,’’ Medical Image Analysis, vol. 36, pp. 61–78, feb 2017. [Online]. Available: https://doi.org/10.1016%2Fj.media.2016.10.004. [12] D. Nie, L. Wang, Y. Gao, and D. Sken, ‘‘Fully convolutional networks for multi-modality isointense infant brain image segmentation,’’ Proc IEEE Int Symp Biomed Imaging, apr 2016. [Online]. Available: https://doi.org/10.1109%2Fisbi.2016.7493515. [13] M. Rajchl, M. C. H. Lee, O. Oktay, K. Kamnitsas, J. PasseratPalmbach, W. Bai, M. Damodaram, M. A. Rutherford, J. V. Hajnal, B. Kainz, and D. Rueckert, ‘‘Deepcut: Object segmentation from bounding box annotations using convolutional neural networks,’’ IEEE TRANSACTIONS ON MEDICAL IMAGING, vol. 36, no. 2, pp. 674–683, 2017. [14] J. Kleesiek, G. Urbana, A. Hubert, D. Schwarz, K. Maier-Hein, M. Bendszus, and A. Biller, ‘‘Deep mri brain extraction: A 3d convolutional neural network for skull stripping,’’ NeuroImage, vol. 129, pp. 460–469, 2016. [15] M. Havaei, A. Davy, D. Warde-Farley, A. Biard, A. Courville, Y. Bengio, C. Pal, P.-M. Jodoin, and H. Larochelle, ‘‘Brain tumor segmentation with deep neural networks,’’ Medical Image Analysis, vol. 35, pp. 18–31, jan 2017. [Online]. Available: https://doi.org/10.1016%2Fj.media.2016.05.004. [16] P. Moeskops, M. A. Viergever, A. M. Mendrik, L. S. de Vries, M. J. N. L. Benders, and I. Isgum, ‘‘Automatic segmentation of mr brain images with a convolutional neural network,’’ IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1252–1261, may 2016. [Online]. Available: https://doi.org/10.1109%2Ftmi.2016.2548501. [17] S. Pereira, A. Pinto, V. Alves, and C. A. Silva, ‘‘Brain tumor segmentation using convolutional neural networks in MRI images,’’ IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1240–1251, may 2016. [Online]. Available: https://doi.org/10.1109%2Ftmi.2016.2538465. [18] Bollmann, Steffen, Andrew Janke, Lars Marstaller, David Reutens, Kieran O’Brien, and Markus Barth. ‘‘MP2RAGE T1-Weighted Average 7T Model,’’ January 1, 2017. doi:10.14264/uql.2017.266. [19] Munk, J., N. Jacobsen, M. Plocharski, L. R. Østergaard, M. Barth, A. Janke, and S. Bollmann. ‘‘Contrast Matching of Ultra-High Resolution Minimum Deformation Averaged MRI Models to Facilitate Computation of a Multi-Modal Model of the Human Brain.’’ In Proc. Intl. Soc. Mag. Reson. Med. 25, 1352. Honolulu, 2017. http://indexsmart.mirasmart.com/ISMRM2017/PDFfiles/1352.html. [20] Janke, Andrew L., and Jeremy F. P. Ullmann. ‘‘Robust Methods to Create Ex Vivo Minimum Deformation Atlases for Brain Mapping.’’ Methods, Spatial mapping of multi-modal data in neuroscience, 73 (February 2015): 18–26. doi: 10.1016/j.ymeth.2015.01.005.

695 Evaluation of 2D texture analysis on fMRI data to identify changes in the striatal network induced by alcohol drinking

S653 Purpose/Introduction: Alcohol use disorders are one of the most prevalent mental health problems and its detection at early stages is crucial. In this study we propose a classification model using 2D texture features in fMRI to discriminate the occurrence of moderate levels of alcohol drinking in subjects according to textural alterations in their striatal network, network considered of relevance in drug addiction1. Subjects and Methods: Eighteen Marchigian-Sardinian rats, a model of chronic excessive alcohol consumption, underwent T2*-weighted resting-state fMRI and T2-weighted anatomic images before (control condition) and after 30 days of alcohol consumption (alcohol condition). Functional images were registered to a rat brain T2-weigthed MRI template to extract the region of interest (ROI). For this purpose, group independent component analysis was performed, using both control and alcohol conditions, identifying different resting-state networks1. Binary mask of striatal network was obtained and analyzed in each of the 36 data-sets. For each ROI, 43 texture features were extracted using first and second-order statistical methods2. Feature values were standardized to zero mean and unit variance. Six predictive models were studied to evaluate their discriminative power. Evaluation was performed using a leave-group-out test. Data was randomly divided into training (75%) and testing (25%) a total of 100 times. Every time, training set was used to build the model and testing set to evaluate the classifier performance. Classification results were provided averaging the area under the ROC curve (AUC) on test sets. Feature selection process was computed using solely training sets. For each partition, a filter feature selection method was applied, obtaining a ranking with the most discriminative texture features. Statistical significance of each feature was evaluated by the filter using the p-value provided by a paired t-test. The whole process was tested for five different graylevel values to evaluate their influence in the final results, since graylevels affect the texture feature computation by modifying the signalto-noise ratio. Results: High classification accuracy (AUC [ 0.895) was obtained for all models. The best classification result (AUC = 0.9325 ± 0.097) was achieved using the Naive Bayes model with a subset of only 7 features and using 32 gray-levels (Figure). The number of gray-levels had important influence in the classification. Discussion/Conclusion: The results reveal that the proposed approach is able to discriminate between both conditions (control and alcohol exposed) with high accuracy. Therefore, changes in striatal network induced by alcohol drinking can be clearly detected using the information extracted from texture analysis, even after a relative short period of exposure. References: 1 Pe´rez-Ramı´rez U., Dı´az-Parra A., Ciccocioppo R., Canals S., Moratal D. ‘‘Brain Functional Connectivity Alterations in a Rat Model of Excessive Alcohol Drinking’’. Conf. Proc. IEEE Eng. Med. Biol. Soc. (EMBC), Jeju, Korea, 2017. 2 Vallie`res M., Freeman C. R., Skamene S. R., El Naqa I. ‘‘A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metastases in soft-tissue sarcomas of the extremities’’. Phys. Med. Biol., vol. 60, pp. 5471–96, 2015.

´ . Pe´rez-Ramı´rez1, A. Dı´azS. Ruiz-Espan˜a1, R. Ortiz-Ramo´n1, U 1 2 3 Parra , R. Ciccocioppo , S. Canals , D. Moratal1 1 Center for Biomaterials and Tissue Engineering, Universitat polite`cnica de Vale`ncia, Valencia/SPAIN, 2School of Pharmacy, University of Camerino, Camerino/ITALY, 3Instituto de Neurociencias, Consejo Superior de Investigaciones Cientı´ficas, Univerdidad Miguel Herna´ndez, Alacant/SPAIN

123

S654

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Number Mean age (sd) #Male/#Female

Subject Statistics Controls Neurodegenerative 13 66 63.3 (9.7) 68.5 (7.6) 0.571 0.625

Two radiologists from the Department of Neuroradiology (Freiburg, Germany) compiled two groups of subjects: patients suffering from neurodegenerative disorders and patients having unimpared perfusion (see Table). The processing described in Figure 1 was done with Matlab and the medical imaging platform OpenNORA ( http://www.nora-imaging.com). A low-dimensional representation was computed by PCA to determine the linear subspace which contains the maximal amount of variance.

696 Dimensionality Reduction Of Arterial Spin Labeling Reveals An Age Gradient J.H. Kirchner1, E. Kaya1, E. Kellner2, S. Yang1, S. Kohl1, H. Urbach1, K. Egger1 1 Dept. of Neuroradiology, Medical Center-University of Freiburg, Faculty of Medicine, University of Freiburg, Germany, Freiburg/ GERMANY, 2Dept. of Radiology, Medical Physics, Faculty of Medicine, Medical Center University of Freiburg, Freiburg/ GERMANY Purpose/Introduction: While Machine Learning techniques have already produced impressive results in a research setting, transfering these techniques into the clinical routine still remains a difficult task. One of the major difficulties relates to the interpretability of high dimensional feature spaces which are commonly used by these algorithms. Principal Component Analysis (PCA) is a tool for reducing high dimensional problems down to a manageable size [1]. In this work I extracted a low dimensional representation from Arterial Spin Labeling (ASL) imaging data for visualization and to facilitate subsequent Machine Learning processes. Subjects and Methods:

Results:

From Figure 3 it is apparent that there is an age gradient along the major axis in PC space. This can be explained by the decreasing amount of perfusion with increasing age [2, 3]. However, the two major PCs which explain more than 80% of the overall variance do not encode perfusion uniformly across the whole brain but predominently concentrate on the classically affected regions in the parietal lobe (Figure 2). A consistent observation was made in [4].

Discussion/Conclusion: The present work confirms the usefulness of

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 ASL for the diagnosis of neurodegenerative diseases. The representations of the PCs in MNI space suggest that the strongest changes in ASL perfusion correlate to the respective neurological changes. We are currently collecting a larger sample to locate these regions with the highest perfusion variability with more confidence. In agreement with the observation made in [3] the age of the subject is a strong predictive factor for neurodegenerative diseases as is suggested by the age gradient in Figure 3. A procedure for the clinical routine must not ignore this fact and should integrate it into its prediction. Finally it must be noted that dimensionality reduction can significantly improve performance of machine learning approaches [1]. The results of this analysis give reason to assume that for Machine Learning problems with ASL perfusion data a significant reduction of the problem size can be achieved in many clinically relevant situations. References: [1] Mwangi, Benson, Tian Siva Tian, and Jair C. Soares. ‘‘A review of feature reduction techniques in neuroimaging.’’ Neuroinformatics 12.2 (2014): 229–244. [2] Biagi, Laura, et al. ‘‘Age dependence of cerebral perfusion assessed by magnetic resonance continuous arterial spin labeling.’’ Journal of Magnetic Resonance Imaging 25.4 (2007): 696–702. [3] Collij, Lyduine E., et al. ‘‘Application of machine learning to arterial spin labeling in mild cognitive impairment and Alzheimer disease.’’ Radiology 281.3 (2016): 865–875. [4] Chen, J. Jean, H. Diana Rosas, and David H. Salat. ‘‘Age-associated reductions in cerebral blood flow are independent from regional atrophy.’’ Neuroimage 55.2 (2011): 468–478.

S655

Clinical Review Posters 697 Epilepsia Partialis continua Multimodality imaging features reexplored Pearls every resident must know R. Vadapalli1, A.S. Vadapalli2, S.J. Sattaluri3, M. Pnigrahi3 1 radiology, Vijaya diagnostics, Hyderabad/INDIA, 2Medicine, AFMC PUNE, pune/INDIA, 3Neuroepileptology, KIMS Hyderabad, Secunderabadrabad/INDIA Introduction: Background: Epilepsia partialis continua (EPC) is characterized by a simple partial motor seizure, restricted to one part of the body with repetitive regular or irregular clonic jerks without loss of consciousness. EPC is presently considered as a type of seizure rather than a form of epilepsy and is categorized under focal status epilepticus. Learning Objectives: To illustrate common and uncommon pathological substrates detected on Imaging in patients with EPC. To discuss the role of multi-modality imaging approach using ictal EEG, 3T high resolution MRI, Ictal SPECT and Inter ictal FDG PET with optimal EPC imaging protocol. Cases: Content organization: Key points on Clinical and Ictal EEG findings of EPC are discussed with common and uncommon epileptogenic substrates listed with illustrative examples. Imaging features lesions causing EPC on MRI, 3D high resolution corticography, Diffusion Tensor imaging, Ictal SPECT and inter-ictal FDG PET are discussed with radio pathological correlation.

123

S656

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

698 Arterial spin labelling: Basics and current emerging clinial applications what every resident must know R. Vadapalli1, A.S. Vadapalli2 1 Radiology, Vijaya diagnostics, Hyderabad/INDIA, 2Medicine, AFMC PUNE, Pune/INDIA

Discussion: Epilepsia partialis continua (EPC) is a rare form of focal status epilepticus commonly involving motor cortex having vascular, immune-mediated, neoplastic or metabolic-toxic causes. This educational exhibit show cases the clinical, EEG, Multi modality Imaging features of EPC. Epilepsia partialis continua can be considered the status epilepticus equivalent of simple partial motor seizures. It manifests itself as focal motor clonic seizures without jacksonian March, seizures remain localized to the part of the body in which they originate, and motor activity is often persistent, lasting for at least 60 min and often for hours, days, weeks, or even longer. Motor activity often is modified by sensory stimuli. Frequency is usually 0.1–6 Hz. Epilepsia partialis continua can continue for long periods (sometimes years) without spreading, although spread can occur at times. Epilepsia partialis continua often is associated with postictal or interictal weakness. Metabolism: The epilepsia partialis continua focus (1) is hypermetabolic, as shown by 2-deoxyglucose positron emission tomography (PET) scanning; (2) is hyperactive, as seen as by magnetoencephalogram (MEG) studies; and (3) shows increased blood flow in single-photon emission computed tomography (SPECT) scan. References: Vein AA, van Emde Boas W. Kozhevnikov epilepsy: the disease and its eponym. Epilepsia. 2011 Feb. 52(2):212–8. . Kojewnikoff AY. Eine besondere Form von corticaler Epilepsie. Neurologie Zentralblatt. 1895. 14:47–48. Bruns. Anastomische und klinische Demostrationen. Neurol ZentraIbI. 1895. 14:47–48. Orlovskij, S. B. Ein Fall von epilepsia partialis continua (russ.) Medskoe. Obozr. Sprimona 43. Neurol. Zbl. 1825. 526:891–892. Choroschko VK. Zur Differentialdiagnostik der Polyclonia epileptoides continua (Kozevnikov). Neurol Zbl. 1908. 27:279. Wilson G, Winkelman NW. Partial Continuous Epilepsy. Arch Neurol Psychiat. 1924. 11:530–42. Spiller WG Martin E. Epilepsia partialis continua occurring in cerebral syphilis. JAMA. 1908. 52:1921. Burr CW. Continnuous Clonic Spasm of the Left A (Epilepsia Continua) Caused by a Tumor of the brain. Amer J med Sci. 1915. 149:169.

123

Introduction: Arterial spin labelling (ASL) is a non-invasive, quantitative, repeatable MRI technique, which allows measuring the brain perfusion without contrast agent using arterial blood water as an endogenous tracer and eliminating the risk of nephrogenic systemic fibrosis in patients with renal dysfunction. A radiofrequency pulse (RF) using to invert the water molecules—this is the labelling part of the ASL. After a delay (so-called post-labelling delay (PLD) or inversion times (TI)) the labelled blood flows into the brain tissue and a labelled image is acquired which contain signal from the inverted- and static tissue protons. A control image is also necessary without labelling. The difference between the control and labelled images provide a measure of labelled blood from arteries delivered to the tissue by perfusion. Cases: Content organization: Key points of types ASL techniques and PCASL (Pseudo continuous ASL} in specific and its pearls and pitfalls are discussed. Present indications and emerging indications are illustrated. Challenges of ASL in clinical reality, artifacts and solutions to avoid them are discussed.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S657 blood flow in whole brain and vascular territories. J Magn Reson Imaging. 2003;18:649–655. Duhamel G, de Bazelaire C, Alsop DC. Evaluation of systematic quantification errors in velocity-selective arterial spin labeling of the brain. Magn Reson Med. 2003;50:145–153. Wang J, Licht DJ. Pediatric perfusion MR imaging using arterial spin labeling. Neuroimaging Clin N Am. 2006;16:149–167. ix.

699 Diffusion-weighted imaging as a simple method for visualization of cerebrospinal fluid dynamics on MRI Discussion: ASL perfusion MRI: basic concept. Arterial blood is labeled or tagged and, after a delay, moves into the imaging plane or volume, during which time there is T1 decay of the label. Snapshot images are acquired in labeled and control conditions and subtracted. There are two major approaches for ASL: continuous (CASL) and pulsed (PASL) In CASL, arterial blood water is continuously and selectively labeled as it passes through a labeling plane, typically applied at the base of the brain. Labeling of all blood water occurs at the same location, and labeling can be applied for several seconds, maximizing the effects on brain signal. In PASL, a short RF pulse is used to instantaneously invert blood and tissue, and can be applied either below the brain, or to the entire brain with subsequent selective inversion of the imaging slices to produce a magnetization difference between blood and brain water. While CASL provides stronger perfusion contrast, it is more difficult to implement due to hardware demands, and deposits a higher level of RF power into the subject compared to PASL. ASL imaging sequences incorporating high field, parallel imaging, pseudo-CASL, and 3D imaging with background suppression now provide an approximately tenfold increase in sensitivity for imaging CBF as compared to prior approaches. Clinical aplications: Acute and chronic cerebrovascular disease. AV malformations. Neoplasms. Epilepsy. Neurodevolopmental disorders. Aging and neurodegenerative disorders. Paediatric cerebral blood flow studies. Perfusion based FMRI. References: Wong EC. Quantifying CBF with pulsed ASL: technical and pulse sequence factors. J Magn Reson Imaging. 2005;22:727–731. Wong EC, Cronin M, Wu WC, Inglis B, Frank LR, Liu TT. Velocityselective arterial spin labeling. Magn Reson Med. 2006;55:1334–1341. Wang J, Alsop DC, Song HK, et al. Arterial transit time imaging with flow encoding arterial spin tagging (FEAST) Magn Reson Med. 2003;50:599–607. Detre JA, Samuels OB, Alsop DC, Gonzalez-At JB, Kasner SE, Raps EC. Noninvasive magnetic resonance imaging evaluation of cerebral blood flow with acetazolamide challenge in patients with cerebrovascular stenosis. J Magn Reson Imaging. 1999;10:870–875. Ances BM, McGarvey ML, Abrahams JM, et al. Continuous arterial spin labeled perfusion magnetic resonance imaging in patients before and after carotid endarterectomy. J Neuroimaging. 2004;14:133–138. Floyd TF, Ratcliffe SJ, Wang J, Resch B, Detre JA. Precision of the CASL-perfusion MRI technique for the measurement of cerebral

E. Yamashita1, T. Yamane1, Y. Tanabe2, T. Ogawa2 1 Division of clinical radiology, Tottori university hospital, Yonago/ JAPAN, 2Division of Radiology, Department of Pathophysiological Therapeutic Science, Faculty of Medicine, Tottori university, Yonago/ JAPAN Introduction: Diffusion-weighted imaging (DWI) provides signal proportional to the water molecular diffusion. It means the extreme water motion-sensitive imaging. In general, higher b values (e.g., b = 1000 s/mm2) are usually used to prevent the effects of moving water molecules. It can sensitively depict disease-associated changes of random translational molecular motion, known as diffusion or Brownian water motion. On the other hand, water molecules with a large degree of motion or a great diffusion distance (e.g., jets of urine flow from the ureter into bladder) will show signal attenuation on DWI with low b values (e.g., b = 10–100 s/mm2) (Fig. 1). Recently, FOCUS-DWI is proposed as a novel method that uses a 2-dimensional spatially selective echo-planar radiofrequency excitation pulse and a 180refocusing pulse reducing the FOV in the phase-encode direction. It can lead to decrease of number of k-space acquisition lines and significantly reduce the imaging distortion (1). The purpose of this study was to develop novel cerebrospinal fluid (CSF) flow cine imaging as a simple method with FOCUS-DWI.

Cases: Firstly, phantoms were scanned to demonstrate the feasibility of FOCUS-DWI and its advantages in improving resolution and reducing distortion. To simulate in vivo behavior of CSF in a phantom, the flow with a pump was produced into a bottle filled with water. The flow cine imaging was obtained by scanning FOCUS-DWI repeatedly without any gating (Fig. 2). Next, we performed CSF flow cine imaging with FOCUS-DWI in 3 volunteers and 5 patients with aqueductal stenosis and hydrocephalus (Fig. 3).

123

S658

Discussion: Although flow cine imaging as CSF dynamics with MRI is obtained by phase contrast method with ECG gating, its acquisition and image analysis are not simple (2). Moreover, CSF dynamics, which are circulated in various types of motion according to respiratory motion as well as cardiac pulsation, have been reported (3). The repetition measurements of FOCUS-DWI allow to obtain a flow cine imaging. It requires no gating and is particularly useful for CSF measurement. Additionally, FOCUS-DWI is capable of providing high resolution images, and can achieve the distortion less than EPIDWI methods. This technique would allow to obtain the flow cine imaging easily and enables to obtain new findings not obtained by conventional methods. References: 1. Feng Z, Min X, Sah VK, Li L, Cai J, Deng M, Wang L. Comparison of field-of-view (FOV) optimized and constrained undistorted single shot (FOCUS) with conventional DWI for the evaluation of prostate cancer. Clin Imaging. 2015; 39(5): 851–5. 2. Wang CS, Wang X, Fu CH, et al. Analysis of cerebrospinal fluid flow dynamics and morphology in Chiari I malformation with cine phase-contrast magnetic resonance imaging. Acta Neurochir (Wien). 2014; 156(4): 707–13. 3. Yamada S, Miyazaki M, Kanazawa H, et al. Visualization of cerebrospinal fluid movement with spin labeling at MR imaging: Preliminary results in normal and pathophysiologic conditions. Radiology. 2008; 249(2): 644–652.

700 Asymptomatic T1 pallidal hyperintensity in patients with chronic liver disease F. Costa, J.P. Filipe, A. Aires, C. Reis, J. Fonseca Neuroradiology, Centro Hospitalar Sa˜o Joa˜o, Porto/PORTUGAL

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Introduction: T1 pallidal hyperintensity is believed to be caused by the deposition of neuroactive substances in the brain, namely ammonia and heavy metals, particularly manganese. Manganese accumulates in a selective manner, with the globus pallidus and substancia nigra being preferred sites for unknown motives [1–6]. These changes may precipitate Acquired Hepatocerebral Degeneration (AHD), a rare chronic encephalopathy that may follow any form of chronic liver disease (CLD) associated with portosystemic shunts. Its clinical spectrum comprises motor signs, cognitive impairment, less frequently short term memory loss and in some cases mielopathy. It is believed that its prevalence (AHD) is approximately from 1 to 2% in patients with CLD [7, 8]. But little is known about the prevalence of T1 pallidal hyperintensity in chronic liver patients population that bare no such symptoms and may therefore be wrongfully labeled as AHD. Cases: We retrospectively analyzed the medical records and imaging studies of CLD patients, for whom brain MRI data was available. The presence of T1 hyperintensity, its location and signal intensity was registered. We used non-parametric tests and Fisher’s exact test to evaluate the association between symptomatic and asymptomatic patient and T1 signal intensity, hyperintensity location and Child– Pugh scores. Two hundred and fifty patients were identified, of which 16 (6.4%) had T1 pallidal hyperintensity, and among them only 7 (2.8%) presented with AHD clinical spectrum symptoms, meaning that 56% of patients with T1 pallidal hyperintensity showed no symptomatology whatsoever. Comparing symptomatic and asymptomatic patients, we observed no significant univariate association between the two groups and Child–Pugh score, signal intensity or location. Discussion: The T1 hyperintense pallidal signal is not predictive of AHD in this exploratory study, raising the question of its exact role in the clinical diagnosis of AHD. References: 1. Meissner W, Tison F. Acquired hepatocerebral degeneration. Handbook of Clinical Neurology 2011; 100(3rd): 193–197. 2. Hauser RA, Zesiewicz TA, Rosemurgy AS, Martinez C, Olanow CW. Manganese intoxication and chronic liver failure. Ann Neurol 1994; 36: 871–875. 3. Spahr L, Butterworth RF, Fontaine S, et al. Increased Blood Manganese in Cirrhotic Patients: Relationship to Pallidal Magnetic Resonance Signal Hyperintensity and Neurological Symptoms. Hepatology 1996;24:1116–1120. 4. Klos KJ, Ahlskog JE, Kumar N, Cambern S, Butz J, Burritt M, Josephs KA. Brain metal concentrations in chronic liver failure patients with pallidal T1 MRI hyperintensity. Neurology 2006; 67(11): 1984–1989. 5. Finlayson M, Superville B. Distribution of cerebral lesions in acquired hepatocerebral degeneration. Brain 1981; 104: 79–95. 6. Butterworth RF. Parkinsonism in cirrhosis: Pathogenesis and current therapeutic options. Metabolic Brain Disease 2013; 28(2): 261–267. 7. Jog MS, Lang AE. Chronic Acquired Hepatocerebral Degeneration: case reports and new insights. Mov Disord 1995; 10: 714–722. 8. Rose C, Butterworths R, Zayed J, et al. Manganese deposition in basal ganglia structures results from both portal-systemic shunting and liver dysfunction. Gastroenterology 1999; 117: 640–644.

701 Gelastic seizures: Multimodality Imaging susbstrates and their pathological correlates,-Pictorial review R. Vadapalli1, A.S. Vadapalli2, S.J. Sattaluri3, M. Panigrahi3

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S659

1

radiology, Vijaya diagnostics, Hyderabad/INDIA, 2Medicine, AFMC PUNE, pune/INDIA, 3Neuroepileptology, KIMS Hyderabad, Secunderabadrabad/INDIA Introduction: Gelastic epilepsy is a rare but well recognized epileptic syndrome typically manifesting in early childhood. It is characterized by recurrent brief seizures with initial laughter or grimacing. It is commonly seen in Hypothalamic Hamartomas but can also be seen in other disorders like Tuberous Sclerosis, temporal lobe lesions, focal obstructive hydrocephalus, focal cortical dysplasia etc. locations other than hypthalamic hamartomas include cinguate gyrus, Inferior frontal gyrus etc. Learning Objectives: Illustrate the common clinical EEG and Multi modality Imaging features in patients with Gelastic seizures wit radiopathological correlation. Cases: Content organization: Common and Uncommon aetiologies of gelastic seizures are discussed with radio pathological correlation by illustrative examples. Multi-modality Imaging features of refractory epilepsies with gelastic seizures on MRI, Ictal SPECT, Inter ictal FDG PET with intra-operative and Histopathological correlation are illustrated. Children with HH present with epileptic attacks of laughter (gelastic seizures) and/or rare episodes of crying (dacrystic seizures). Hypothalamic gelastic epilepsy appears to be extremely rare, probably 0.1% of all patients with seizures. Surgical removal/disconnection of HH may suppress gelastic seizures. Efficacy and safety of different surgical approaches (e.g. stereotactic, endoscopic, radiosurgical) require further study.

Discussion: Gelastic seizures are epileptic events characterized by bouts of laughter. Laughter-like vocalization is usually combined with facial contraction in the form of a smile. Autonomic features such as flushing, tachycardia, and altered respiration are widely recognized. Gelastic seizures have been associated classically to hypothalamic hamartomas, although different extrahypothalamic localizations have been described. Hypothalamic hamartomas are rare congenital lesions presenting with the classic triad of gelastic epilepsy, precocious puberty and developmental delay. The clinical course of patients with gelastic seizures associated with hypothalamic hamartomas is progressive, commencing with gelastic seizures in infancy, deteriorating into more complex seizure disorder resulting in intractable epilepsy. Electrophysiological, radiological, and pathophysiological studies have confirmed the intrinsic epileptogenicity of the hypothalamic hamartoma. Currently the most effective surgical approach is the trancallosal anterior

123

S660

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

interforniceal approach, however newer approaches like endoscopic radiosurgery and gamma knife have been used with success. GS are associated with the Pallister Hal syndrome (PHS). This syndrome is an autosomal dominant inheritable disorder characterized by HHs, polydactyly, laryngeal malformations such as bifid epiglottis, pulmonary segmentation anomalies, imperforate anus, and panhypopituitarism. References: Sethi PK, Rao ST. Gelastic, quiritarian, and cursive epilepsy: a clinicopathological appraisal. J Neurol Neurosurg Psychiatry. 1976;39:823–8. Shin HY, Hong SB, Joo EY, et al. Gelastic seizures involving the right parietal lobe. Epileptic Disord. 2006;8:209–12. Maixner W. Hypothalamic hamartomas—clinical, neuropathological and surgical aspects. Childs Nerv Syst. 2006;22:867–73. Mathieu D, Kondziolka D, Niranjan A, et al. Gamma knife radiosurgery for refractory epilepsy caused by hypothalamic hamartomas. Stereotact Funct Neurosurg. 2006;84:82–7. McConachie NS, King MD. Gelastic seizures in a child with focal cortical dysplasia of the cingulate gyrus. Neuroradiology. 1997;39:44–5. Mohamed IS, Otsubo H, Shroff M, et al. Magnetoencephalography and diffusion tensor imaging in gelastic seizures secondary to a cingulate gyrus lesion. Clin Neurol Neurosurg. 2007;109:182–7.

702 Imaging biomarkers of Hirayama disease: Conventional dynamic MRI and DTI findings of Hirayama’s Disease: Pictorial review R. Vadapalli1, A.S. Vadapalli2, R.D. Mulukutla3, S. Vemula4 1 radiology, Vijaya diagnostics, Hyderabad/INDIA, 2Medicine, AFMC PUNE, pune/INDIA, 3SPINE AND ORTHOPEDIC SURGERY, UDAI OMNI HOSPITALS, hYDERABAD/INDIA, 4neurology, Magna hospitals, Hyderabad/INDIA Introduction: First described by Hirayama et al. in 1959, sporadic juvenile muscular atrophy of the upper limb affects young men in period of rapid growth predominantly called as Hirayama’s disease. It is a very rare benign lower motor neuron disorder, also referred to as monomelic amyotrophy (MMA). It is characterized by the insidious onset and progressive weakness and wasting of a distal extremity. Pathophysiology: In the normal spine, the spinal dura is attached at two places—one at the foramen magnum, C2 and C3, and the other at the coccyx—and is anchored to the vertebral canal at the nerve root exits. In healthy subjects, this dura is slack and loosely suspended and consists of several transverse folds, which compensate for the increased length of the cervical canal in flexion. As against this, in Hirayama disease, a short dura cannot compensate for the increased length in flexion and so is displaced anteriorly, with resultant compression of the spinal cord. The clinical features include insidious onset, predominantly unilateral upper extremity weakness and atrophy, cold paresis and no sensory or pyramidal tract involvement. The amyotrophy is unilateral in most patients, asymmetrically bilateral in some and rarely symmetric in others. Cases: Conventional MRI, Dynamic Flexion MRI and Diffusion tensor imaging based morphological and microstructural biomarkers provide diagnostic and prognostic characterization of Hirayama’s disease.

123

Discussion: Conventional MRI with Neutral position and Flexion Imaging biomarkers: Localized atrophy, abnormal cervical curvature, asymmetrical cord flattening, loss of normal Dural attachment, anterior of dorsal dura on flexion, deformed thecal sac on flexion, engorged posterior epidural venous plexus with thoracic extension with abnormal peri medullary flow voids. Diffusion tensor Imaging Biomarkers (in neutral and Flexion positions): Significant reduction of Fractional Anisotropy values (FA values) in the posterior [ anterior cord with reduced fibre density in posterior columns with reduced colour saturation (Blue colour) on colour coded directionality DTI maps. References: Hirayama K. Non-progressive juvenile spinal muscular atrophy of the distal upper limb [Hirayama’s disease] In: De Jong JM, editor. Handbook of clinical neurology. Vol. 15. Amsterdam, the Netherlands: Elsevier; 1991. pp. 107–20.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S661

Sonwalkar HA, Shah RS, Khan FK, Gupta AK, Bodhey NK, Vottath S, et al. Imaging features in Hirayama disease. Neurol India. 2008;56:22–6. Biondi A, Dormont D, Weitzner I, Jr, Bouche P, Chaine P, Bories J. MR imaging of the cervical cord in juvenile amyotrophy of distal upper extremity. AJNR Am J Neuroradiol. 1989;10:263–8. Gourie-Devi M, Suresh TG, Shankar SK. Monomelic amyotrophy. Arch Neurol. 1984;41:388–94. Pradhan S, Gupta RK. Magnetic resonance imaging in juvenile asymmetric segmental spinal muscular atrophy. J Neurol Sci. 1997;146:133–8. Tashiro K, Kikuchi S, Itoyama Y, Tokumaru Y, Sobue G, Mukai E, et al. Nationwide survey of juvenile muscular atrophy of distal upper extremity (Hirayama disease) in Japan. Amyotroph Lateral Scler. 2006;7:38–45. Chen CJ, Chen CM, Wu CL, Ro LS, Chen ST, Lee TH. Hirayama Disease: MR Diagnosis. AJNR Am J Neuroradiol. 1998;19:365–8. Hirayama K, Tomonaga M, Kitano K, Yamada T, Kojima S, Arai K. Focal cervical poliopathy causing juvenile muscular atrophy of distal upper extremity: A pathological study. J Neurol Neurosurg Psychiatry. 1987;50:285–90.

703 Many Multi-modality imaging Avatars of Vascular dementia and its mimics and variants: what every neuro resident must know R. Vadapalli1, A.S. Vadapalli2, S.J. Sattaluri3 1 Radiology, Vijaya diagnostics, Hyderabad/INDIA, 2Medicine, AFMC PUNE, Pune/INDIA, 3Neuroepileptology, KIMS Hyderabad, Hyderabad/INDIA Introduction: Vascular Dementia represents a complex dementia subtype that occurs as a result of interaction between vascular risk factors, infarcts, white matter changes and brain atrophy. Sub types of VaD (Cortical infarcts, SIVD (subcortical ischemic vascular disease) and small vessel disease (SVD), Cerebral microbleeds and amyloid angiopathy/. Clinical stages of vascular dementia are: Stage 1: No impairment. Stage 2: Very mild decline. Stage 3: Mild decline. Stage 4: Moderate decline (mild or early stage). Stage 5: Moderately severe decline (moderate or mid-stage). Stage 6: Severe decline (moderately severe or mid-stage). Stage 7: Very severe decline (severe or late stage). Cases: – Clinical Diagnostic Criteria of vascular dementia and sub types of VaD and their Imaging correlates with pathophysiology are Illustrated with case based examples. – MR Imaging Features, DTI and FDG PET CT features of Vascular dementia are illustrated with differential diagnosis. Atrophy patterns, Ischemia patterns and dentato rubro pallido luysian axis signal abnormalities.

123

S662 Discussion: Imaging subtypes of vascular dementia include the following.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 enabling proper discrimination of most of the letters and digits by the time of the study.

– Strategic infarct related dementia e.g. thalamic or eloquent area infarct. – Single lacunar or multi infarct dementia. – Small vessel disease and binswanger disease with deep whitematter ischemia. – Association of Alzhiemer’s disease as both coexist. – Cerebral microangiopathy-hypertensive type with mltiple cerebral microbleeds in central capsuloganglionic and brainstem distribution. – Amyloid angiopathy with peripheral distribution of cerebral microbleeds. Strategic infarcts and small vessel disease. Cognitive dysfunction in VaD can be the result of. 1. Large vessel infarctions: Bilateral in the anterior cerebral artery territory. Parietotemporal- and temporo-occipital association areas of the dominant hemisphere (angular gyrus included). o Posterior cerebral artery territory infarction of the paramedian thalamic region and inferior medial temporal lobe of the dominant hemisphere. Watershed infarctions in the dominant hemisphere (superior frontal and parietal). 2. Small vessel disease: Multiple lacunar infactions in frontal white matter ([ 2) and basal ganglia ([ 2). WMLs (at least more than 25% of WM). Bilateral thalamic lesions. References: Hachinski VC, Bowler JV. Vascular dementia. Neurology. 1993 Oct. 43(10):2159–60; discussion 2160–1.

704 Compensatory role of the cerebellum? - An alexia partial recovery case report P. Naumczyk1, A. Marcinkowska2, A. Sabisz3, M. Łockiewicz1, K. Kluj-Kozłowska4, E. Naro_zan´ska5, M. Sildatke-Bauer6, J. Sławek7, E. Szurowska3, E. Sitek7 1 Social Sciences, Institute of Psychology, University of Gdansk, Gdansk/POLAND, 22nd Radiology Department, Medical University of Gdansk, Gdansk/POLAND, 3Second Department of Radiology, Medical University of Gdansk, Gdansk/POLAND, 4Department of Polish Philology, University of Gdansk, Gdansk/POLAND, 5 Neurology Department, Copernicus, Szpital sw. Wojciecha, Gdansk/ POLAND, 6Ophthalmology Department, Copernicus, Szpital sw. Wojciecha, Gdansk/POLAND, 7Department of Nursing, Medical University of Gdansk, Gdansk/POLAND Introduction: This is to report a case of a 64-year-old patient with diagnosed alexia without agraphia. We aimed at exploring brain mechanisms underlying the improvement in patient’s reading performance following neuropsychological rehabilitation. Cases: In 2010 the patient being reported had an ischemic stroke localized within the medial occipital and posterior temporal cortex of the left hemisphere as well as the right cerebellum cortex. The injury was followed by cortical atrophy of left occipital lobe (see Fig. 1). Four years after the incident, the patient started neuropsychological rehabilitation being diagnosed with alexia, slight anomia and optical aphasia. After three years patient’s reading performance improved

123

The MRI examination was held on 1.5T Siemens Aera with 20-channel head coil. It included T1-MPR sequence (TR 1730 ms, TE 3.3 ms, 1 mm3 voxel, 176 axial slices), DTI sequence (TR 11000 ms, TE 85 ms, 2 mm3 voxel, 76 axial slices, b = 1000, 30 directions, NEX 3), and two functional T*2-EPI sequences requiring reading of familiar and novel words (TR 3000 ms, TE 45 ms, voxel 3.5 9 3.5 9 4.5 mm, 32 AC-PC slices, 220 and 330 dynamic scans respectively). Fiber tracts were reconstructed on the Siemens Syngovia workstation with the default processing parameters (tract length of 50–400 mm, 2 seeds per voxel, angle and FA threshold of 30o/0.2). Upon the reconstruction the Inferior Longitudinal Fasciculus (ILF) in each hemisphere was labeled basing on the Wakana (2007) specification. The reconstruction is presented on Fig. 2, the statistics of the tracts— in Table 1. Tab.1 Statistics of the reconstructed Inferior Longitudinal Fasciculus (ILF) tracts of both hemispheres ILF - left hemisphere ILF - right hemisphere Number of tracts 250 513 Mean FA 0.48 0.44 Mean Diffusivity 0.0009 0.0008 Mean RD 0.0006 0.0006 Mean Ad 0.014 0.0013

The anatomical and functional images were processed in the

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 FreeSurfer package (Fischl, 2012). The Functional Analysis Stream (FS-FAST) was applied to the fMRI data, including slice timing, motion correction and spatial smoothing (5 mm FWHM). The results of the analyses covering the cerebellum are presented on Fig. 3.

Discussion: We focused on the ILF as the tract connects regions associated with reading skills (Catani, 2008). Patient’s left reconstructed ILF was significantly smaller than the right hemisphere one. It also consisted of an additional collateral fiber aiming the right cerebellum cortex, which corresponded with clusters of cerebellum activation during reading. This suggests cerebellum involvement in the partial recovery of reading skills of the patient. References: Catani M, de Schotten MT. (2008). A diffusion tensor imaging tractography atlas for virtual in vivo dissections. Cortex, 44, 1105–1132. Fischl, B. (2012). FreeSurfer. NeuroImage, 62(2), 774–781. Wakana, S. et al. (2007). Reproducibility of quantitative tractography methods applied to cerebral white matter. Neuroimage, 36, 630–644.

705 Carcinoid tumor of the middle ear: a potential misdiagnosis of cholesteatoma on HASTE DWIs M. Ersen, H.T. Sanal, B. Arik, M. Tasar Radiology, Gu¨lhane Training and Research Hospital, Ankara/ TURKEY Introduction: Carcinoid tumors of the middle ear are exceedingly rare masses. They cause challenges in the differentiation of middle ear tumors. In this case report, we aimed to present a carcinoid tumor of the middle ear which may lead to misdiagnosis of cholesteatoma owing to their identical imaging findings. Cases: 50 year-old female patient with hearing loss, treated conventionally for chronic otitis media in an outer center 3 years ago. The patient applied to otorhinolaryngology department since the complaints did not relieve despite antibiotherapy. Computerized tomography (CT) of the left temporal bone revealed a soft tissue density lesion in the middle ear cavity destroying bony chain and infiltrating mastoid cells. On HASTE diffusion weighted magnetic resonance imaging (MRI), diffusion restriction found to be compatible with cholesteatoma was reported. With conventional MRI sequences, the lesion was slightly hyperintense compared to cerebrospinal fluid on T1-W and markedly hyperintense on T2-W images. Enhancement was not seen on post-contrast series. The histopathological evaluation was reported as a carcinoid tumor. Discussion: Carcinoid tumors may be accepted as neuroendocrine adenomas of middle ear. Although some sources consider the adenomas and carcinoid tumors separate entities, recent studies report

S663 these lesions as adenomas showing distinctive neuroendocrine differentiation. These rare tumors may be misdiagnosed as cholesteatoma because of their diffusion restriction and other imaging characteristics identical to these lesions. Thus, neuroendocrine tumors, although rare, must be kept in mind in the differential diagnosis of HASTE diffusion restricted middle ear lesions. References: 1. Hasan, Zubair, et al. ‘‘Neuroendocrine Adenoma of the Middle Ear: A Rare Histopathological Diagnosis.’’ Case Reports in Otolaryngology 2016 (2016). 2. Ramsey, Mitchell J., et al. ‘‘Carcinoid tumor of the middle ear: clinical features, recurrences, and metastases.’’ The Laryngoscope 115.9 (2005): 1660–1666. 3. Torske, Kevin R., and Lester DR Thompson. ‘‘Adenoma versus carcinoid tumor of the middle ear: a study of 48 cases and review of the literature.’’ Modern pathology 15.5 (2002): 543–555.

706 Differentiation between papillary renal cell carcinoma and fat-poor angiomyolipoma: a preliminary study assessing detection of intratumoral hemorrhage with chemical shift MRI and T2*-weighted gradient echo S.Y. Kim, S. Woo, J.Y. Cho, S.H. Kim Radiology, Seoul National University Hospital, SEOUL/KOREA, REPUBLIC OF Introduction: Objective: To determine whether intratumoral hemorrhage detected using chemical shift MRI and T2*-weighted gradient echo (GRE) can be used to differentiate papillary renal cell carcinoma (pRCC) from fat-poor angiomyolipoma (fpAML). Materials and Methods: This retrospective study included 42 patients with pRCC (n = 28) or fpAML (n = 14) who underwent MRI followed by surgery. Two blinded radiologists independently assessed the presence of intratumoral hemorrhage using chemical shift MRI (decrease in signal intensity from opposed- to in-phase) and T2*-weighted GRE (‘‘blooming’’). Consensus reading was determined for discrepant cases. MRI findings were compared using Chi square test. Inter-observer agreement was assessed using kappa statistics. Results: Inter-observer agreement was substantial for both sequences (k = 0.622 and 0.793, p \ 0.001). For chemical shift MRI, the prevalence of intratumoral hemorrhage was significantly greater in pRCC than in fpAML (71.4 vs 28.6%, p = 0.019 for reader 1; 64.3 vs 14.3%, p = 0.003 for reader 2; and 75 vs 21.4%, p = 0.002 for the consensus). T2*-weighted GRE showed a similar tendency, with a significantly greater prevalence of intratumoral hemorrhage in pRCC than in fpAML (46.4 vs 14.3%, p = 0.049 for both readers; and 50 vs 14.3%, p = 0.042 for the consensus). When using the consensus reading, the sensitivity and specificity of determining pRCC were 75 and 78.6% for chemical shift MRI and 50 and 85.7% for T2*weighted GRE. Conclusion: We found that prevalence of loss of SI from the T2* effect using chemical shift MRI and T2*-weighted GRE was significantly different between pRCC and fpAML. Two hemorrhage sensitive MR sequences reflect a greater prevalence of hemorrhage for pRCC than for fpAML. Although both MRI sequences demonstrated substantial inter-observer agreement, there was a trend for better agreement when using T2*-weighted GRE. These hemorrhagesensitive MRI sequences may be used as an adjunctive tool for discriminating between the two entities.

123

S664 Cases:

We found that prevalence of loss of SI from the T2* effect using chemical shift MRI and T2*-GRE was significantly different between pRCC and fpAML. Two hemorrhage sensitive MR sequences reflect a greater prevalence of hemorrhage for pRCC than for fpAML. Discussion: In the current study, we evaluated two hemorrhagesensitive MRI sequences for the differentiation between pRCC and fpAML. The results of our study demonstrated that intratumoral hemorrhage on MRI was significantly more common in pRCC compared with fpAML using both sequences of chemical shift MRI and T2*-weighted GRE. As the differentiation between pRCC and fpAML is often challenging in clinical practice, we believe that detection of intratumoral hemorrhage using both chemical shift MRI and T2*-weighted GRE can

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 potentially be used as adjunctive tool for differentiating between the two entities. Both sequences were categorized as showing substantial inter-observer agreement (0.622 and 0.793 for chemical shift MRI and T2*weighted GRE, respectively). The inter-observer agreement may be greater using T2*-weighted GRE than using chemical shift MRI. References: 1. Bissler JJ, Kingswood JC. Renal angiomyolipomata. Kidney international 2004; 66:924–934. 2. Jinzaki M, Silverman SG, Akita H, Nagashima Y, Mikami S, Oya M. Renal angiomyolipoma: a radiological classification and update on recent developments in diagnosis and management. Abdom Imaging 2014; 39:588–604. 3. Woo S, Cho JY, Kim SH, Kim SY. Angiomyolipoma with minimal fat and non-clear cell renal cell carcinoma: differentiation on MDCT using classification and regression tree analysis-based algorithm. Acta Radiol 2014; 55:1258–1269. 4. Sasiwimonphan K, Takahashi N, Leibovich BC, Carter RE, Atwell TD, Kawashima A. Small (\ 4 cm) renal mass: differentiation of angiomyolipoma without visible fat from renal cell carcinoma utilizing MR imaging. Radiology 2012; 263:160–168. 5. Schieda N, Dilauro M, Moosavi B, et al. MRI evaluation of small (\ 4 cm) solid renal masses: multivariate modeling improves diagnostic accuracy for angiomyolipoma without visible fat compared to univariate analysis. Eur Radiol 2016; 26:2242–2251. 6. Hindman N, Ngo L, Genega EM, et al. Angiomyolipoma with minimal fat: can it be differentiated from clear cell renal cell carcinoma by using standard MR techniques? Radiology 2012; 265:468–477. 7. Amin MB, Corless CL, Renshaw AA, Tickoo SK, Kubus J, Schultz DS. Papillary (chromophil) renal cell carcinoma: histomorphologic characteristics and evaluation of conventional pathologic prognostic parameters in 62 cases. Am J Surg Pathol 1997; 21:621–635. 8. Hakim SW, Schieda N, Hodgdon T, McInnes MD, Dilauro M, Flood TA. Angiomyolipoma (AML) without visible fat: Ultrasound, CT and MR imaging features with pathological correlation. Eur Radiol 2016; 26:592–600. 9. Yoshimitsu K, Kakihara D, Irie H, et al. Papillary renal carcinoma: diagnostic approach by chemical shift gradient-echo and echo-planar MR imaging. J Magn Reson Imaging 2006; 23:339–344. 10. Childs DD, Clingan MJ, Zagoria RJ, et al. In-phase signal intensity loss in solid renal masses on dual-echo gradient-echo MRI: association with malignancy and pathologic classification. AJR Am J Roentgenol 2014; 203:W421–428. 11. Murray CA, Quon M, McInnes MD, et al. Evaluation of T1Weighted MRI to Detect Intratumoral Hemorrhage Within Papillary Renal Cell Carcinoma as a Feature Differentiating From Angiomyolipoma Without Visible Fat. AJR Am J Roentgenol 2016; 207:585–591. 12. Lin DD, Filippi CG, Steever AB, Zimmerman RD. Detection of intracranial hemorrhage: comparison between gradient-echo images and b(0) images obtained from diffusion-weighted echo-planar sequences. Am J Neuroradiol 2001; 22:1275–1281. 13. Hsu RM, Chan DY, Siegelman SS. Small renal cell carcinomas: correlation of size with tumor stage, nuclear grade, and histologic subtype. AJR Am J Roentgenol 2004; 182:551–557. 14. Chavhan GB, Babyn PS, Thomas B, Shroff MM, Haacke EM. Principles, techniques, and applications of T2*-based MR imaging and its special applications. Radiographics 2009; 29:1433–1449. 15. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159–174. 16. Patterson J, Lohr D, Briscoe C, Briscoe G, Flanigan RC. Calcified renal masses. Urology 1987; 29:353–356. 17. Chen CL, Tang SH, Wu ST, et al. Calcified, minimally fat-contained angiomyolipoma clinically indistinguishable from a renal cell carcinoma. BMC Nephrol 2013; 14:160.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S665

18. Sun J, Xing Z, Xing W, et al. Intratumoral Macroscopic Fat and Hemorrhage Combination Useful in the Differentiation of Benign and Malignant Solid Renal Masses. Medicine (Baltimore) 2016; 95:e2960. 19. Yu S, Qiu J, Zhang J, Pan L, Xing S, Zhang L. Detection of intratumoral susceptibility signals using T2*-weighted gradient echo MRI in patients with clear cell renal cell carcinoma. PloS One 2013; 8:e79597. 20. Chen J, Ding J, Dai Y, et al. Assessment of intratumoral micromorphology for patients with clear cell renal cell carcinoma using susceptibility-weighted imaging. PloS One 2013; 8:e65866. 21. Chen J, Sun J, Xing W, et al. Prediction of nuclear grade of clear cell renal cell carcinoma with MRI: intratumoral susceptibility signal intensity versus necrosis. Acta Radiol 2014; 55:378–384.

707 Pictorial Review: Magnetic Resonance Imaging Findings of Left Ventricle Hypertrabeculation and Noncompaction F.G. Rodrı´guez-Ruiz, L. Rodrı´guez-Ortiz, J. Maldonado-Vargas Department of Diagnostic Radiology, University of Puerto Rico School of Medicine, San Juan/PUERTO RICO Introduction: Cardiac noncompaction is an uncommon type of primary cardiomyopathy distinguished by excess trabeculation of the myocardium and extensive intertrabecular recesses [1]. Cardiac magnetic resonance imaging (CMRI) is the gold standard for confirmation of diagnosis due to greater differentiation between trabeculated and compacted myocardium, better apex evaluation and use of contrast agents [2]. MRI diagnostic criteria for cardiac noncompaction proposed by Petersen et al. [3] is as follows: appearance of a compacted epicardial layer and a noncompacted endocardial layer, trabeculations and deep intertrabecular recesses within the noncompacted layer and end-diastole noncompacted-to-compacted myocardial ratio greater than 2.3. Cases: Normal Cardiac Study:

Discussion: The aim of this pictorial review is to discuss CMRI findings of cardiac noncompaction seen in patients from our Institution. This pictorial review will include the normal imaging appearance of the cardiac anatomy to understand imaging findings of cardiac noncompaction. Additionally, an overview of pathophysiology, clinical manifestations, progression, and complications will be discussed. Knowledge of radiological features is essential for prompt diagnosis and adequate timely management of patients. References: [1] Kawel-Boehm N, McClelland RL, Zemrak F, Captur G, Hundley WG, Liu CY, Moon JC, Petersen SE, Ambale-Venkatesh B, Lima JAC, et al. 2017. Hypertrabeculated left ventricular myocardium in relationship to myocardial function and fibrosis: the multi-ethnic study of atherosclerosis. Radiology. 0:0. [2] Zuccarino F, Vollmer I, Sanchez G, Navallas M, Pugliese F, Gayete A. 2015. Left ventricular noncompaction: imaging findings and diagnostic criteria. Am J Roentgenol. 204:W519-W530. [3] Petersen SE, Selvanayagam JB, Wiesmann F, Robson MD, Francis JM, Anderson RH, Watkins H, Neubauer S. 2005. Left ventricular non-compaction: insights from cardiovascular magnetic resonance imaging. J Am Coll Cardiol. 46(1):101–105.

123

S666

708 MR Imaging of the complications in hepatic hydatid disease D. Herek Radiology, Pamukkale University School of Medicine, Denizli/ TURKEY Introduction: Hydatid disease is a zoonotic infection caused by the larval stage of the different types of Echinococcus cestode, mostly the type: Echinococcus granulosus. Humans are accidental hosts and are not involved in the natural life cycle of the parasite (1–3). Infection with Echinococcus granulosus affects almost every organ in human body but most affected organs are reported to be liver and lungs (3–5). Radiologically ultrasound (US), computed tomography (CT) and magnetic resonance imaging (MRI) are used in detection of the disease itself and the complications if exist. Ultrasound easily shows the membranes, septa and sand (4). Computed tomography is useful in the detection of calcifications. Magnetic resonance imaging is the best choice for showing the features detected by US besides complications as cysto-biliary and cysto-pleural fistula, rupture, infected cyst and for planning surgery and interventional therapy (1, 4, 7, 8). In this report, we aim to present some complications of hepatic hydatid disease and emphasize the importance of MRI in the diagnosis of complications. Cases:

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Discussion: Magnetic resonance imaging is the best radiologic modality in the detection of primary and recurrent hydatid disease and for diagnosis of complications such as super infection, contained and direct rupture into peritoneal cavity, biliary system or pleural cavity resulting in cysto-biliary and cysto-pleural fistula (7, 8). References: ¨ ztekin, MP. Akdogan. 1) AS. Eksioglu, B. Ucan, E. C¸akmakcı, PS. O Imaging Findings in Liver Hydatidosis: Pictorial Essay. ECR 2014 Congress, poster no:1790, DOI:10.1594/ecr2014/C-1790. 2) M. Rodrı´guez Parodi, J. Alvarez, C. Bianco, V. Ortiz Imaging findings in vertebral hydatid cyst: contribution of MRI and PET CT. ECR 2017 Congress, poster no: C-0883, doi:10.1594/ecr2017/C-0883. 3) Herek D, Karabulut N. CT demonstration of pulmonary embolism due to the rupture of a giant hepatic hydatid disease. Clin Imaging. 2012 Sep-Oct;36(5):612–4. 4) Pedrosa I, Saı´z A, Arrazola J, Ferreiro´s J, Pedrosa CS. Hydatid disease: radiologic and pathologic features and complications. Radiographics. 2000 May-Jun;20(3):795–817. 5) Dahniya MH, Hanna RM, Ashebu S, Muhtaseb SA, el-Beltagi A, Badr S, el-Saghir E. Hydatid Cyst: A Pictorial Review of Radiological Appearances. Br J Radiol 2001 Mar;74(879):283–9. 6) Polat P, Atamanalp SS. Hepatic Hydatid Disease: Radiographics Findings. Eurasian J Med. 2009 Apr;41(1):49–55. 7) Polat P, Kantarci M, Alper F, Suma S, Koruyucu MB, Okur A. Hydatid Disease from Head to Toe. Radiographics. 2003 MarApr;23(2):475–94; quiz 536–7. 8) Mehta P, Prakash M, Khandelwal N. Radiological manifestations of hydatid disease and its complications. Trop Parasitol. 2016 JulDec;6(2):103–112.

709 Development and optimisation of a 3T MRI protocol for Use in oesophageal cancer staging research K. Foley1, P. Griffiths2, A. Roberts3, A. Riddell4 1 School of Medicine, Cardiff University, Division of Cancer & Genetics, Cardiff/UNITED KINGDOM, 2Institute of Life Sciences, Swansea University, Swansea/UNITED KINGDOM, 3Department of Radiology, University Hospital of Wales, Cardiff/UNITED KINGDOM, 4Department of Radiology, Royal Marsden Hospital, Surrey/UNITED KINGDOM Introduction: Oesophageal cancer (OC) has a poor prognosis and is currently staged with CT, endoscopic ultrasound (EUS) and positronemission tomography (PET). Recent evidence suggests that the diagnostic accuracy of N-staging with CT, EUS and PET is low. (Foley, 2017) In addition, patients with oesophageal tumours too stenotic to allow passage of the endoscope are reliant on CT and PET, which are known to under-stage disease. High-resolution T2 MRI sequences have shown promising results for local staging (Riddell, 2007) and early research studies using ex vivo specimens have shown excellent accuracy with higher strength magnets (7T). (Yamada, 2015) This exploratory study investigates 3T MRI in healthy volunteers and explores differences in image quality, aiming to develop and optimise a novel protocol for future OC staging research studies. Cases: Four healthy volunteers gave consent to participate (3 males, 1 female, median age 38.5 years, range 29–46). Two different T2weighted sequences were acquired and evaluated for image quality; a T2 half-fourier-acquired single-shot turbo spin echo (HASTE) sequence and a TSE sequence with radial blades, a movement correction tool developed by Siemens Healthcare, named BLADE. A radiologist assessed the image quality considering the resolution, signal-to-noise ratio (SNR), ability to visualise important anatomical

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S667

landmarks, and scored the images accordingly. (Fig. 1) The T2 axial HASTE sequences with larger pixel and voxel size were considered better image quality due to reduced noise artefact. (Table 1) The T2 BLADE sequences were considered inadequate quality. Table 1. Image Parameters for T2 HASTE sequences

TR (ms) 1730 1730 1730 1730

TE (ms) 96 96 96 96

Matrix Size 156x256 194x320 194x380 234x384

FOV (mm) 255x340 240x320 285x350 255x340

Slice Thickness (mm) 4 4 4 3

Pixel Size (mm) 1.33 1.00 0.92 0.89

Voxel Size (mm3) 7.06 4.00 3.39 2.35

Score 4 4 3 2

Discussion: Oesophageal imaging with 3T MRI is feasible and important anatomical landmarks required for OC staging can be visualised. T2 HASTE sequences have been developed and optimised for this purpose. Further validation of 3T oesophageal imaging is required. References: Foley KG, et al. Accuracy of contemporary oesophageal cancer lymph node staging with radiological-pathological correlation. Clinical Radiology. 2017. In press. http://dx.doi.org/10.1016/j.crad.2017.02.022. Riddell AM, et al. The appearances of oesophageal carcinoma demonstrated on high-resolution, T2-weighted MRI, with histopathological correlation. European Radiology. 2007;17(2):391–399. Yamada I, et al. Ultra-high-resolution MR imaging of esophageal carcinoma at ultra-high field strength (7.0 T) ex vivo: correlation with histopathologic findings. Magnetic Resonance Imaging. 2015;33(4):413–419.

710 Pictorial Review: MRI evaluation of perianal fistulas in crohn’s disease L. Rodrı´guez-Ortiz, L. Figueroa-Diaz, M. Betancourt-Torres, J. Lara-Rı´os, G. Ballester-Ortiz Department of Diagnostic Radiology, University of Puerto Rico School of Medicine, San Juan/PUERTO RICO Introduction: Approximately one quarter of patients with Crohn’s disease commonly present with perianal fistulas (fistulas-in-ano) [1], which represent abnormal connections between the anal canal and the perineal skin. Knowledge of anal canal anatomy (Fig. 1a) is mandatory for identification and adequate classification of perianal fistulas. Fistulas in this anatomical region are reported according to the anal clock (Fig. 1b). Additionally, their classification is based on the Parks system [2], which group such fistulas according to their relation to sphincteric muscles (Fig. 1c). Evaluation with magnetic resonance imaging (MRI) plays an essential role in determining location in relation to pelvic floor structures, as well as identifying common complications and planning surgical management.

Cases: The following are cases of perianal fistulas in patients diagnosed with Crohn’s Disease:

123

S668

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Discussion: The aim of this pictorial review is to discuss the classification system used to evaluate the anal canal and present radiological features and complications of perianal fistulas as seen in cases of patients with Crohn’s disease. References: [1] Sheedy SP, Bruining DH, Dozois EJ, Faubion WA, Fletcher JG. 2017. MR imaging of perineal Crohn Disease. Radiology. 282(3):628–645. [2] Parks AG, Gordon PH, Hardcastle JD. 1976. A classification of fistula-in-ano. Br J Surg. 63: 1–12.

711 Utility of PET/MR in head and neck cancer T. Sekine1, F. Barbosa2, G. Delso2, E. Ter Voert2, P. Veit-Haibach2, M. Huellner2 1 Department of Radiology, Nippon Medical School, Tokyo/JAPAN, 2 Department of Nuclear Medicine, University Hospital Zurich, Zurich/SWITZERLAND Introduction: The combined MRI components on PET/MR system serve detailed anatomy which could impact on the diagnosis of head and neck cancer. The aim of this presentation is to understand the rolle of PET/MR in the assessment of head and neck cancer by comparing it with PET/ CT. Cases: The contents of the case presentation is as below. 1. 2. 3. 4. 5. 6.

T staging. N staging. Tumor response. Resectability-defining findings. Perineural spreading. Artifact.

Discussion: Compared with PET/CT, PET/MR is subtle superior in T staging and in the assessment of resectability. In the lesions with dental artifact, PET/MR could serve more precise images than PET/CT.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 PET/MR is at least comparable to PET/CT in the assessment of head and neck tumor. Good workflow on PET/MR study is important in the clinical setup. References: 1. Sekine T, Barbosa FG, Delso G, et al. Local resectability assessment of head and neck cancer: Positron emission tomography/MRI versus positron emission tomography/CT. Head Neck 2017: 2. Sekine T, Barbosa F, Kuhn FP, et al. PET + MR versus PET/CT in the initial staging of head and neck cancer, using a trimodality PET/ CT + MR system. Clin Imaging 2017.

712 WITHDRAWN

713 How to evaluate prostate MRI in 7 steps S¸ .B. Arik, M. Tasar, M. Ersen, H.T. Sanal, K.N. Arda, B. Karaman, S. Hamcan, U. Bozlar Radiology, Gu¨lhane Training and Research Hospital, ANKARA/ TURKEY Introduction: Today Magnetic Resonance Imaging (MRI) is the most favorable non-invasive evaluation modality in prostate imaging. Owing to the fast advances in coil technology and new sequences, different kinds of opinions in clinical approach and follow up of prostate cancer cases may arise and hence, the improvements in MRI technology can also cause discordances among clinicians patient evaluations. In this review, we aim to display the steps in the evaluation of prostate MRI referring to ACR PI-RADS v2 guide. Cases: Step1: 1st Case: Anatomy; Normal, PI-RADS 1. Step2: Patient Preparation. Step3: MRI Technique. Step4: PI-RADS v2 criteria. 2nd Case: PI-RADS 2 Peripheral gland lesion. 3rd Case: PI-RADS 2 Central gland lesion. 4th Case: PI-RADS 3 Central gland lesion, DWI negative (final score PI-RADS 3). 5th Case: PI-RADS 3 Peripheral gland lesion, DCE positive (final score PI-RADS 4). 6th Case: PI-RADS 4 Central gland lesion (erased charcoal). 7th Case: PI-RADS 5 Peripheral gland lesion. 8th Case: PI-RADS 5 Peripheral gland lesion with extra capsular extension. Step 5: Additional Findings not mentioned in PI-RADS v2. 9th Case: Positive lymph node (internal iliac). 10th Case: Bone metastasis. Discussion: Step 6: PI-RADS restrictions. Step 7: Why perform MRI, before biopsy. References: Weinreb, J.C., et al., PI-RADS Prostate Imaging—Reporting and Data System: 2015, Version 2. Eur Urol, 2016. 69(1): p. 16–40. Panebianco, V., et al., Pitfalls in Interpreting mp-MRI of the Prostate: A Pictorial Review with Pathologic Correlation. Insights Imaging, 2015. 6(6): p. 611–30. Rosenkrantz, A.B. and S.S. Taneja, Radiologist, be aware: ten pitfalls that confound the interpretation of multiparametric prostate MRI. AJR Am J Roentgenol, 2014. 202(1): p. 109–20. Panebianco, V., et al., Multiparametric magnetic resonance imaging vs. standard care in men being evaluated for prostate cancer: a randomized study. Urol Oncol, 2015. 33(1): p. 17 e1–7.

S669 Talab, S.S., et al., Prostate cancer imaging: what the urologist wants to know. Radiol Clin North Am, 2012. 50(6): p. 1015–41. Yacoub, J.H., A. Oto, and F.H. Miller, MR imaging of the prostate. Radiol Clin North Am, 2014. 52(4): p. 811–37. Shaish, H., S.S. Taneja, and A.B. Rosenkrantz, Prostate MR Imaging: An Update. Radiol Clin North Am, 2017. 55(2): p. 303–320.

714 Cardiac MRI: examination and evaluation steps with case examples U. Bozlar1, M. Tasar1, M. Ersen2, B. Arik2, S. Hamcan2, K.N. Arda2, H.T. Sanal2 1 Radiology, Gu¨lhane Training and Research Hospital, ANKARA/ TURKEY, 2Radiology, Gu¨lhane Training and Research Hospital, Ankara/TURKEY Introduction: Cardiac Magnetic Resonance (MR) Imaging is a useful and problem-solving tool in evaluating cardiac diseases, contributing to high soft tissue resolution and cine images. Unfamiliar imaging planes and cine images show difference from other MR scans and cause difficulties in evaluating. In this pictorial, we tried to perform cardiac MR with routine sequences and pathological case examples in 10 steps. Cases: Step I: CMR imaging planes: -Case I—Healthy individual. Step II: Basic CMR sequences with samples. -Case II—Healthy individual. Step III: Evaluation of chambers size and wall-thickness. -Case III—Healthy individual. -Case IV—Dilated Cardiomyopathy. -Case V—Hypertrophic cardiomyopathy (HCM). Step IV: Evaluation of global function. -Case VI—Ejection fraction (EF) calculation. -Case VII—Dilated Cardiomyopathy. Step V: Evaluation of focal wall motion abnormalities. -Case VIII—Healthy individual. -Case IX—Arrhythmogenic Right Ventricular Cardiomyopathy (ARVC). Step VI: Evaluation of heart valves. -Case X—Patient with, cine views. -Case XI—Systolic anterior motion (SAM) of mitral valve. -Case XII—Mitral valve prolapse. Step VII: Evaluation of atrial and ventricular cavities: thrombus, mass etc. -Case XIII—Atrial thrombus. -Case XIV—Calcified Amorf Tumor. Step VIII: Evaluation of late gadolinium enhancement (LGE). Case XV—LGE acute myocarditis. Case XVI—LGE in chronic MI. Case XVII—LGE in HCM. Step IX: Evaluation of additional sequences for specific lesions. Case XIX—Myocardial tagging. Step X: Evaluation of adjacent structures (Pericardial effusion, aberrant sol SVC). Case XX—Pericardial effusion. Case XXI—Persistent left superior vena cava. Discussion: Cardiac MRI is a required modality for a wide variety of cardiac pathologies like congenital anomalies, masses, myocardial ischemia, cardiomyopathies etc. Not containing ionizing radiation, better soft tissue resolution and obtainable additional sequences are main superiorities of cardiac MRI. When cardiac MR is assessed by adhering to certain principles, more meaningful results can be

123

S670

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

obtained. Detailed clinical information ant the patient’s previous examinations, plays an important role in the approach. References: 1. Ginat, Daniel T., et al. ‘‘Cardiac imaging: part 1, MR pulse sequences, imaging planes, and basic anatomy.’’ American Journal of Roentgenology 197.4 (2011): 808–815. 2. Rajiah, Prabhakar. ‘‘Cardiac MRI: part 2, pericardial diseases.’’ American Journal of Roentgenology 197.4 (2011): W621-W634. 3. Gaba, Ron C., et al. ‘‘Cardiovascular MR Imaging: Technique Optimization and Detection of Disease in Clinical Practice 1.’’ Radiographics 22.6 (2002): e6-e6.

715 Experience with 3T, 32-channel Head Coil and Multi Transmit in the diagnosis of Malignant Melanomas in the eye H. Simonsen1, K. Segers2, B. Moeller2 1 Functional Imaging Unit, Dept. of Clinical Physiology, Nuclear Medicine and PET, Rigshospitalet, Glostrup, Glostrup/DENMARK, 2 Departepartepartment of Radiology, Rigshospitalet, Glostrup, Glostrup/DENMARK Introduction: Melanomas are mostly found on the skin but also occur inside the eye. Frequently they appear as brown spots on the iris, but can also be found in the beam body and the choroid. It is not known why this kind of malignance of the eye occurs, but it is not due to irradiance with ultra violet light. In Denmark the number has remained constant at just under 50 cases per year over the past several years. Five to ten patients get melanomas of the iris, while the other occur in the beam body and especially in the choroid. Patients with melanoma of the eye have previously been studied with CT and Ultrasound. [1]. Upgrading our MR system to 32-channel head coil and MultiTransmit parallel RF transmission, allows us to scan faster and get higher spatial and temporal resolution. An increase in SNR means that we can scan with small field of view over the eye and still get high resolution images. Cases: The MRI examination is performed on a 3 T Philips Achieva (Philips Healthcare, The Netherlands) scanner, equipped with an 32-element receive head coil and MultiTransmit parallel RF transmission. Patients are placed comfortably in the centre of the coil and are asked to keep their eyes closed and rested.

Discussion: Using a 32-channel head coil instead of a dedicated coil for eye examination allows us to make a diagnostic examination of the whole head without changing the coil and still get images with high SNR in a small FOV. One of the major challenges when scanning the eyes is to obtain images without eye movement. We have discussed whether a fixation point could help, but we think that the patients will get tired and start to blink. Instead we ask the patients to close their eyes and keep them rested. To help patients with difficulties to keep their eyes still, we sometime put a cloth over their eyes. MultiTransmit parallel RF transmission helps us to increase the image homogeneity also in areas near air-filled spaces as we have around the eyes. Conclusion: With our optimized protocol we have shown that the clinicians can get valuable information about the location, size and placement of the melanoma and during the same examination is able to identify if the patient has other diseases in the brain. References: [1] Ondartede svulster i øjets indre. Professor Jan Ulrik Prause http://vos.dk/37043VOS16.html.

716 Novel multimodal diagnostic approach for patients with chronic disorders of consciousness E. Kremneva, L. Legostaeva, E. Mochalova, S. Morozova, D. Sinitsyn, D. Sergeev, A. Poydasheva, O. Chervyakova, Y. Ryabinkina, N. Suponeva, M. Piradov Radiology, Research Center of neurology, Moscow/RUSSIAN FEDERATION Introduction: Chronic disorders of consciousness (DOC), such as VS and MCS, are diagnosed mostly by clinical examination, that carries high risk of misdiagnosis. New tools may help establish level of consciousness in certain cases, as presented here. Cases: A clinically VS patient A., male, 48 y.o., 2 years after intracerebral hemorrhage (CRS-R = 10) showed only reflex movements to stimulation, no environment contact and preserved sleep– wake pattern. It was reported recently that presence of default mode network (DMN) signal on rs-fMRI was correlates with the degree of clinical consciousness impairment (Vanhaudenhuyse, 2010). On rsfMRI (3T) we found activation of DMN areas (mPFC, PCC, lIPC,

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S671

rIPC) (Fig. 1). We also performed TMS-EEG with calculating Perturbational Complexity Index (PCI)—an independently validated promising consciousness metric (Casali, 2013), that allows reliable stratification of unresponsive patients with empirical cutoff level for discrimination between the unconscious and conscious states of 0.31 (Casarotto, 2016). We found out high complexity of the cortical response for the TMS stimuli, with PCI of 0.345 for frontal region stimulation (Fig. 2) and 0.424 for parietal region (Fig. 3) stimulation which implies the «conscious» state in this patient.

Discussion: Novel diagnostic techniques may reveal patients with possible higher level of consciousness than seen clinically. Such patients should become subject for further investigation to find out the cause of discrepancy between clinical and neurophysiological results, as well as for intensive rehabilitation interventions. The study is supported by Russian Scientific Foundation grant §1615-00274. References: 1. Vanhaudenhuyse A1, Noirhomme Q, Tshibanda LJ et al. Default network connectivity reflects the level of consciousness in noncommunicative brain-damaged patients. Brain. 2010 Jan; 133(Pt 1):161–71. doi:10.1093/brain/awp313. 2. Casali AG1, Gosseries O, Rosanova M et al. A theoretically based index of consciousness independent of sensory processing and behavior. Sci Transl Med. 2013 Aug 14;5(198):198ra105. doi: 10.1126/scitranslmed.3006294. 3. Casarotto S, Comanducci A, Rosanova M et al. Stratification of unresponsive patients by an independently validated index of brain complexity. Ann Neurol. 2016 Nov;80(5):718–729. doi: 10.1002/ana.24779.

717 Decreased level of consciousness in thalamic hemorrhage patient E. Kremneva1, L. Legostaeva2, E. Mochalova3, A. Poydasheva2, S. Morozova1, D. Sergeev3, D. Sinitsyn2, Y. Ryabinkina3, O. Chervyakova2, N. Suponeva2, M. Piradov3 1 Radiology, Research Center of neurology, Moscow/RUSSIAN FEDERATION, 2Neurorehabilitation, Research Center of neurology, Moscow/RUSSIAN FEDERATION, 3Intensive care unit, Research Center of neurology, Moscow/RUSSIAN FEDERATION

123

S672

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

Introduction: Widespread loss of cerebral connectivity is assumed to underlie the failure of brain mechanisms that support communication and purposeful behaviour following severe brain injury. Recent studies showed preservation of large-scale cerebral networks in minimally conscious state (MCS) patients (Adams, 2000). We present stroke patient who was in MSC after hypertensive thalamic hemorrhage without diffuse cortical damage (Fig. 1) with patterns observed on resting-state fMRI and TMS-EEG similar to consciousness state.

Cases: Patient B., female, 31 y.o. had 18 CRS-R score 1 year after left thalamic hemorrhage with fluctuating consciousness level. She improved gradually with spontaneous eye opening and simple commands following although communication was limited to yes/no answers on voices recognition. We used rs-fMRI to investigate the default mode network (DMN; Vanhaudenhuyse, 2010) residual signal and found out activation in precuneus, bilateral temporo-parietal junctions and medial prefrontal cortex that was similar to healthy controls (Fig. 1). TMS-EEG showed high perturbational complexity index (PCI; Casali, 2013) of 0.383 ([ 0.31, which is considered as cutoff level between unconscious and the conscious conditions; Casarotto, 2016), that corresponds to high complexity of TMS cortical response, same as in healthy controls (Fig. 2, 3).

123

Discussion: By means of novel methods of consciousness detection, based on assessment of cortical activity and cerebral networks, patient appears to be conscious. We hypothesize that isolated dysfunction of thalamo-cortical interactions caused by thalamic damage lead to clinically seen reduced level of consciousness. The study is supported by RSF grant §16-15-00274. References: 1. Adams, J.H., Graham, D.I. and Jennett, B. (2000) The neuropathology of the vegetative state after acute insult. Brain, 123: 1327–1338. 2. Vanhaudenhuyse A1, Noirhomme Q, Tshibanda LJ et al. Default network connectivity reflects the level of consciousness in noncommunicative brain-damaged patients. Brain. 2010 Jan; 133(Pt 1):161–71. doi: 10.1093/brain/awp313.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S673

3. Casali AG1, Gosseries O, Rosanova M et al. A theoretically based index of consciousness independent of sensory processing and behavior. Sci Transl Med. 2013 Aug 14;5(198):198ra105. doi: 10.1126/scitranslmed.3006294. 4. Casarotto S, Comanducci A, Rosanova M et al. Stratification of unresponsive patients by an independently validated index of brain complexity. Ann Neurol. 2016 Nov;80(5):718–729. doi: 10.1002/ana.24779.

719 The use of 3D FLAIR sequence in subarachnoid hemorrhage detection in sedated pediatric patients with acute TBI

718 Failed Back Surgery Syndrome: a pilot study of preliminary chronic pain patterns in fMRI data

Introduction: Despite the fact that the CT is the gold standard imaging method for the detection of acute subarachnoid haemorrhage, it is known that FLAIR sequence in MRI possesses exceptional abilities in detecting subarachnoid hemorrhage of various etiologies. However, CSF flow artifacts and CSF hyperintensity in cerebral sulci of sedated pediatric patients remain the problem for conventional 2D FLAIR and can produce false-positive results. SAH is known to be the one of the strongest predictors of TBI outcome in children. We compared two MRI pulse sequences and CT for subarachnoid hemorrhage detection in acute traumatic brain injury. Cases: We retrospectively evaluated the neuroimaging data of 38 pediatric patients (20 males/18 females, aged from 6 to 16 years with median age 11.6 years) who admitted to our institute with acute severe traumatic brain imaging. CT and MRI in all patients was performed in first 24 h from the moment of injury (with an interval of less than 10 h between the 2 procedures). MRI in all patients was performed under anesthesia with sevoflurane and 60–100% oxygen due to severity of their clinical condition. We evaluated sensitivity and specificity for 2D and 3D FLAIR images in relation to CT data, which were considered the gold standard. The results for 2D FLAIR images in our study were unsatisfactory. Discussion: The paramagnetic effects of supplemental O2 administration result in shortened CSF T1. Radiologists should be aware of this phenomenon to avoid attributing increased CSF signal intensity on FLAIR images to abnormal CSF properties such as hemorrhage or elevated protein content.1 Some authors mention that in sedated pediatric brains, 3D FLAIR almost fully suppresses CSF signal, and enables reliable assessment free from CSF artifacts.2 Our data confirm that the use of 2D FLAIR sequence for detecting subarachnoid hemorrhage in sedated pediatric patients is not advisable. At the same time, the use of 3D FLAIR technique helps to detect acute SAH with considerable sensitivity and specificity (in relation to CT data), which in turn potentially allows a more accurate prediction of the brain injury outcome. The first limitation of our study is that CT and MRI were not conducted at one time. The second limitation is the lack of direct verification of the presence of SAH with lumbar puncture in these patients. References: 1) Deliganis A.V., Fisher D.J., Lam A.M., Maravilla K.R. Cerebrospinal fluid signal intensity increase on FLAIR MR images in patients under general anesthesia: the role of supplemental O2. Radiology. 2001 Jan;218(1):152–6. 2) Ozcan U.A., Isik U., Ozpinar A., Baykan N., Dincer A. Assessment of sedated pediatric brain with 3D-FLAIR sequence at 3T MRI. Brain Dev. 2015 May;37(5):495–500.

L. Piliponis1, J. Usˇinskien_e2, G. Terbetas3, G. Kazakevicˇiu¯t_eJanusˇkevicˇien_e4 1 Faculty of Medicine, Vilnius University, Vilnius/LITHUANIA, 2 Department of Radiology, National Cancer Institute, Vilnius/ LITHUANIA, 3Department of Neurosurgery, Republican Vilnius University Hospital, Vilnius/LITHUANIA, 4Department of Graphical Systems, Vilnius Gediminas Technical University, Vilnius/ LITHUANIA Introduction: Failed Back Surgery Syndrome (FBSS) is a condition of persisting low back pain with or without lumbosacral radiculopathy after 1 or more spine surgeries. Main preoperative risk factors are depression, [than 6 months continuous pain and preexisting neuropathic pain. Spine surgery in chronic low back pain has poor results because of the central sensitization at the level of spinal cord and the ,,pain matrix‘‘of the central nervous system. The main clinical application of pain fMRI would be to contraindicate surgery as a treatment method for patients with altered pain perception and functional brain activity. Cases: To investigate the pain functional consequences, we recorded resting state fMRI followed by block design pain stimulation in FBSS patient with chronic low back pain and neurotypical control who matched in age and sex. Pain stimulus was given via clip compressing distal phalanx of right-hand index finger. The fMRI was performed with GE Optima 450w 1.5 T scanner. fMRI images were analysed with SPM12. We studied the difference in pain activations and resting state between both participants. The threshold for functional activity was selected as 30 voxels (p = 0.001). Pain activations showed differences in FBSS patient: anterior and posterior right insular cortices (IC), both supplementary motor areas (SMA), left primary visual cortex (V1) and both sides of the occipital middle-inferior gyri. Discussion: The diagnostic approach of chronic pain patterns using fMRI could help preventing FBSS. With further larger case–control study research we seek to develop an evident model of chronic pain activation for diagnostic purposes. References: 1. Shapiro CM. The Failed Back Surgery Syndrome. Phys Med Rehabil Clin N Am. 2014; 25(2): 319–40. 2. Apkarian AV, Baliki MN, Geha PY. Towards a theory of chronic pain. Prog Neurobiol. 2009; 87(2): 81–97. 3. Hashmi JA, Baliki MN, Huang L et al. Shape shifting pain: chronification of back pain shifts brain representation from nociceptive to emotional circuits. Brain 2013 Sep; 136(9): 2751–2768.

I. Melnikov, M. Ublinskiy, T. Akhadov Radiology, Clinical and Research Institute of Emergency Pediatric Surgery and Trauma, Moscow/RUSSIAN FEDERATION

123

S674

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

720 1H RMN spectroscopy in vitro in in vivo conditions of different drug A. Heintz, O. Seloi, J.L. Schmidt, A. Chamkhi, S. Potier, C. Fournaize, D. Lambal, M. Lefranc, J. Peltier, C. Desenclos, A. Coutte, R. Popoff, B. Chauffert, M. Boone, P. Votte, J.-M. Constans Somme, CHU Amiens Picardie, Saloue¨l/FRANCE Introduction: Magnetic resonance spectroscopy shows the chemical composition of a tissue or a solution studied (1). The objective of this thesis is to study the possible relationships between the chemical composition of drugs and their effects at the tissue level in patients. Cases: It was important to agree on a drug research protocol in order to replicate the tests on both the same drug at different time intervals but also be able to replicate them on other drugs. To do this, it was necessary to agree on important points to check with each acquisition: The MRI antenna used: micro antenna 47 mm. The temperature: in the MRI tunnel (between 20 and 22 C) and sample. The volume and concentration of drugs to be analyzed: sufficient to have an optimal S/N ratio. The size of the voxel (minimum 240 mm3) and the position of the the measure voxel (at the maximum of the signal to avoid interference with the container) in order to optimize the S/N. Different TE (Echo Time) allowing us to detect the coupled metabolites and to distinguish them from noise for acquisitions with low S/N. We studied three drugs: TEMOZOLOMIDE (2, 3, 4) (used in glioblastomas STUPP chemotherapy/radiotherapy protocol) and RETROVIR (HIV) because they have very good cerebral diffusion and are very often used in protocols. The software used to process the spectra is jMRUI (v5.2) (5), open source processing software. For these two drugs, no post treatment was necessary.

123

Discussion: The first results show that the composition of the drug can be studied in vitro under such in vivo conditions. Several resonances appearing in the spectrum with reproducible positions.

Several over drugs have already been studied but the number of atomic groups is more important and to verify their detection requires more time. This will allow us to study, at the tissue level, the effects of these drugs and be able to try to study the relationship between the spectral and metabolic changes in tissue tumors in different drug metabolites. For some drugs, we will have to study their product of transformations (6). We will then be able to compare the results of the in vitro drug in in vivo condition with the results on the tissues of the patients before and during the treatment. References: 1. R. A. DE GAAF (2007) In Vivo NMR Spectroscopy Principles and Techniques 2nd edition. 2. E. S. NEWLANDS et al. (1997) Temozolomide: a review of its discovery, chemical properties, pre-clinical development and clinical trials. Cancer Treatment Reviews 23, 35–61. 3. A. PACE et al. (2003) Temozolomide chemotherapy for progressive low-grade glioma: clinical benefits and radiological response. Annals of Oncology 14: 1722–1726. 4. R. STUPP et al. (2005) Radiotherapy plus Concomitant and Adjuvant Temozolomide for Glioblastoma The new england journal of medicine 352;10. 5. D. STEPHAN et al. (2009) Quantitation of magnetic resonance spectroscopy signals: the jMRUI software package Meas. Sci. Technol. 20 104035 (9 pp).

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 6. M.MUJAHID et al. (2012) An alternate synthesis of enantiomerically pure levetiracetam (Keppra) Tetrahedron: Asymmetry 23 1512–1515.

721 Glioblastoma reccurence therapies response assessment by spectroscopic and MRI volumetric measurements

S675 (choline + lactate/NAA) could be a relevant judgment criterion to better study the aggressiveness of the tumor, survival prediction and the pathological processes. Other patients will be included in the study wich will give results statistically more significant. Spectroscopic ratios vary earlier than volumetry and give more information on tumoral processes and are sometimes predictive of survival. And it is one of the reason that MRS sequences are useful in patient follow-up.

J.P. Chombar1, A. Heintz1, O. Seloi1, M. Boone1, R. Hanafi1, H. Deramond1, W. Dou2, S. Metembou1, P. Toussaint1, C. Desenclos1, A. Coutte1, M. Lefranc1, D. Le Gars1, B. Chauffert1, J.-M. Constans1 1 Somme, CHU Amiens Picardie, Saloue¨l/FRANCE, 2Electronics, Tsinghua University, Beijing/CHINA Introduction: Glioblastoma (GBM) is the most common malignant primary brain tumor in adults, with severe prognosis (survival median is 14–16 months (1, 2, 3)) the current standard of care is maximal safe surgical resection followed by adjuvant concomitant chemoradiotherapy and subsequent consolidation chemotherapy, generally with temozolomide. Spectroscopy is increasingly used to better understand (GBM) tumor metabolism and the ratios studied allow in some cases a better prediction, for the recurrence (4), of the survival, notably certain ratios such as: Cho/Cr, NAA/Cr, Lac/Cr. The ability to combine all these ratios can allow us to obtain an additional survival index to be taken into account. This is why our study allows us to evaluate whether the Magnetic Resonance Specstoscopy (MRS) coupled with volumetry is predictive of patient survival. Cases: Population: 30 Patients with glioblastoma all biopsied and treated with the STUPP protocol and then bevacizumab (18) with recurrence over a period of 35 months. Patients divided into three different groups according to the type of excision (subtotal (17), partial (7), biopsy (6) and treatment. MRI sequences every 2 months (3 and 1.5 T): diffusion, T2, T2 *, FLAIR, 3D FLAIR, 3D T1 with gadolinium injection; Monovoxel spectroscopy (TE at 35 and 144 ms, sometimes 288 ms). The contrast enhancement, necrosis and FLAIR volumes variations (calculated by Aw.Server) over time are evaluated according to the RANO criteria.

References: 1. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, et al. Radiotherapy plus con-comitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005; 352(10):987–96. doi:10.1056/NEJMoa043330 PMID: 15758009. 2. Gilbert MR, Wang M, Aldape KD, Stupp R, Hegi ME, Jaeckle KA, et al. Dose-dense temozolomide for newly diagnosed glioblastoma: a randomized phase III clinical trial. J Clin Oncol. 2013; 31(32):4085– 91. doi: 10.1200/JCO.2013.49.6968 PMID: 24101040; PubMed Central PMCID: PMC3816958. 3. Dolecek TA, Propp JM, Stroup NE, Kruchko C. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2005–2009. Neuro Oncol. 2012; 14 Suppl 5: v1–49. doi: 10.1093/neuonc/nos218 PMID: 23095881; PubMed Central PMCID: PMC3480240. 4. N.A. Sibtain, F.A. Howe, D.E. Saunders The clinical value of proton magnetic resonance spectroscopy in adult brain tumours. Clinical Radiology (2007) 62, 109–119.

722 Suspicious cardiac MRI findings: cardiac myxoma K. Blank1, N.R. Valeviciene2 1 Faculty of Medicine, Vilnius University, Vilnius/LITHUANIA, 2 Center of Radiology and Nuclear medicine, Vilnius University Hospital Santaros Klinikos, Vilnius/LITHUANIA

The MRS ratios (calculated by jMRUI) that we used in the clinic are Cho/Cr (tumor proliferation), NAA/Cr (infiltration) and Lac/Cr (glycolytic metabolism). Longitudinal statistical analyzes of volumes and ratios. Discussion: One of the results is that glycolytic metabolism and/or proliferation persists despite the absence of contrast enhancement, hyper-perfusion and little variation in volume. The global index

Introduction: Cardiac myxomas are the most common benign primary cardiac tumor presenting in adulthood. Myxoma’s clinical presentation is determined by its location, size and mobility. Usually patients present with one or more symptoms from the triad: embolism, intracardiac obstruction and constitutional symtpoms. The purpose of this case series is to review MR imaging performed in Vilnius University Hospital Santaros Clinics for patients with suspected myxoma and to compare the results with literature. Cases: 23 cardiac myxomas were reviewed retrospectively for MR imaging findings. In this series 23 patients (16 female, 7 male; age range 30–85 years; mean age 60 years) had 14 (60.87%) left atrial, 5 (20.83%) right atrial, 4 (17.39%) right ventricular myxomas. In 17

123

S676 (73.91%) cases the tumor was connected to the septum, 11 (47.83%) had a pedicle. The size of tumor ranged from 7 9 6 mm to 89 9 49 mm. 20 (86.96%) myxomas were singular, 2 (8.70%) were oval, 6(26.09%) polypoid. Hemorrhage was noted in 4 cases (17.39%), 5 (21.74%) tumors caused valve obstruction. 4 cases (36.36%) were described as homogeneous and 7 (63.64%) as heterogenous intensity. In T1 sequence 4 (40.00%) tumors had hypointensive signal, there were 3 (30.00%) tumors of isointensive and 3 (30.00%) of hyperintensive signal. In T2 sequence 1 (16.67%) tumor had hypointensive, 3 (50.00%) isointensive and 2 (33.33%) hyperintensive signal. 16 (80.00%) tumors accumulated contrast medium, 13 (81.25%) of which—non homogenously. Calcifications were present in 3 (13.04%) cases. 11 (47.83%) patients were operated and in 10 (90.91%) cases histological findings confirmed the diagnosis of cardiac myxoma. Discussion: Cardiac MRI is an important diagnostic tool for the evaluation of cardiac myxoma. In literature up to 75% of myxomas are localized in left atrium, while in our study less tumors where found in this chamber and more tumors were diagnosed on the right side of the heart. These results could be explained by the fact that all patients with myxoma in our Clinics had a transthoracic or transesophageal echocardiography performed and just more complicated or debatable cases had undergone a cardiac MRI. References: 1. M.L.Grebenc et al., ‘‘Cardiac Myxoma: Imaging Features in 83 Patients,’’ Radiographics, pp. 673–89, 2002. 2. T.Masui et al. ‘‘Cardiac Myxoma: Identification of Intratumoral Hemorrhage and Calcification on MR Images,’’ American Journal of Roentgenology, pp. 850–852, 1994. 3. R. Cohen et al., ‘‘Atrial Myxoma: A Case Presentation and Review,’’ Cardio Res, pp. 41–44, 2012.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 sequences performed at different time point after the implantation of PP meshes in rats. Following the optimization of the photoinsertion reaction parameters an homogenous coverage of the polymer surfaces with Gd-complexes was obtained allowing a strong T1 signal enhancement. In vitro MRI were done at 7 T as a proof of concept and at 9.4 T with different acquisitions protocols to establish the better parameters for in vivo MRI (Figure 1). Multislices MRI T1 weithed gradien echo sequences with an inversion recovery were selected to be clinically relevant and avoid too long acquisition time. For the validation of the method, commercial PP meshes were modified before intramuscular implantation in a rat model. We follow up the evolution of the mesh signal during 3 weeks (Figure 2).

723 Magnetic resonance investigation of photochemically modified polymer surfaces and their application to MRI-visible medical devices M. Cardoso1, A. Schulz2, L. Allegre2, J. Coudane2, C. Goze-Bac3, B. Nottelet2 1 BioNanoNMRI facility, University of Montpellier, Montpellier/ FRANCE, 2IBMM, Universite´ de Montpellier, Montpellier/FRANCE, 3 Laboratoire Charles Coulomb Plateforme BioNanoNMRI, University of Montpellier, Montpellier/FRANCE Introduction: Direct and stable surface modification of polymers can confer specific properties to medical devices. It is a major target for medical innovation particularly in medical imaging. 1,2 To avoid the use of costly equipments and pre-activation of the target, UV-activated aryl-azide photoinsertion (clip reaction) appears as a powerful approach. However it has been mainly reported for model surface functionalization3 and scarcely for the modification of medical devices.4–5 Preclinical investigation in MRI is an essential step to consider prior to successful industrial transfert. In this contribution, we evaluate with the help of MR imaging and relaxometry, the induced modifications of the relaxing properties of the tissue in the presence of implanted meshes. Cases: DTPA and DOTA ligands have been functionalized with arylazide (clip) moieities prior to complexation with GdCl36H2O. The resulting complexes were sprayed over the surface of polymeric substrates (PP hernia meshes). UV-activation was then carried out to covalently immobilize the clip-agent on the surface. MRI-visibility was evaluated in vitro in gel at 7 and at 9.4 T. Then in vivo visibility and stability was evaluated by T1-weighted MRI gradient echo

123

Discussion: This study presents the MRI investigation used to follow the MRI-clip. Despite some dependence over the polymer nature, in vivo visualization was obtained using a multispray approach allowing for the covalent immobilization of Gd-complexes. This first report on MRI-clip photoinsertion demonstrates the potential of this approach that may offer news opportunities towards surface modification of medical devices for contrast MRI implant. References: 1. A. Teo, A. Mishra, I. Park, Y-J. Kim, W-T. Park, Y-J. Yoon ACS Biomater. Sci. Eng.2016, 2, 454. 2. B. Nottelet, V. Darcos, J. Coudane Eur. J. Pharm. Biopharm. 2015, 97, 350. 3. H. Wang et al. ACS Appl. Mater. Interfaces, 2011, 3, 3463. 4. E. Ge´rard et al. J Polym Sci A Polym Chem 2011, 49, 5087. 5. A. Zhu, M. Zhang, J. Wua, J. Shen Biomaterials 2002, 23, 4657.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

724 Diagnosis of atrial myxomas by dynamic cardiac MR S. Mehra Radiodiagnosis, Pgimer, Dr. Ram Manohar Lohia Hospital, New Delhi/INDIA Introduction: Atrial myxomas are polypoid intracavitary masses that can cause cardiac obstrucrtive symptoms or embolus generation.The clinical presentation being highly nonspecific, imaging has a major role to play in the diagnosis. Cardiac MR is an extremely good diagnostic technique as it provides complete assessment of cardiac morphology and detect mass within the cardiac chambers effectively . Cases: Case 1. A 50 year old male presented with dyspnoea of 3 months duration. The EKG findings were normal. Echocardiography demonstrated nodular thickening adjacent to the inter atrial septum. Cardiac MR study depicted mass in the left atrium prolapsing through the AV valve. Case 2. A 41 year old male presented with history of chest pain and palpitations on exertion of 5 months duration. The EKG findings were normal. Echocardiography demonstrated mass within the left atrium and patient was referred for cardiac MRI. Case 3. A 9 year old boy presented with dyspnoea and orthopnoea of 1 year duration. A cardiac MR was asked for by the cardiologist as the echocardiography revealed mass in the atrium which was superbly depicted and characterized on dynamic MR. Case 4. A 55 year male presented with orthopnoea and palpitations on exertion. Cardiac MR was performed as 2 dD echocardiography demonstrated right atrial mass. A right atrial myxoma was diagnosed on imaging findings on dynamic MR . Discussion: The characteristic signal intensity, location, enhancement pattern of the lesion on dynamic cardiac MR enabled effective and accurate diagnosis of atrial myxomas in all patients referred to us with clinical suspicion. All myxomas demonstrated heterogenous to homogenous enhancement. The T1 signal intensity was variable but T2 signal intensity was increased in all myxomas. We could differentiate myxomas from intra atrial thromus based on the typical features. References: Freedberg RS, Kronzon I, Rumancik WM, Lie-beskind D. The contribution of magnetic resonance imaging to the evaluation of intracardiac tumors diagnosed by echocardiography. Circulation 1988; 77:96–103. Didier D, Ratib O, Friedli B, et al. Cine gradient-echo MR imaging in the evaluation of cardiovascular diseases. RadioGraphics1993; 13:561–573. Araoz PA, Eklund HE, Welch TJ, Breen JF. CT and MR imaging of primary cardiac malignancies. RadioGraphics 1999; 19:1421–1434.

S677

Software Exhibits 725 L1-LAD: Iterative MRI reconstruction using L1 constrained least absolute deviation J.-M. Lin1, H.-C. Chang2, T.-C. Chao3, S.-Y. Tsai4, A. Patterson5, H.-W. Chung6, J. Gillard1, M. Graves5 1 Department of Radiology, University of Cambridge, Cambridge/ UNITED KINGDOM, 2Department of Diagnostic Radiology, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong/HONG KONG PRC, 3Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan/ TAIWAN, 4Department of Applied Physics, National Chengchi University, Taipei/TAIWAN, 5MRIS Unit, Cambridge University Hospitals NHS Foundation Trust, Cambridge/UNITED KINGDOM, 6 Electrical Engineering, National Taiwan University, Taipei/ TAIWAN Purpose of the software: Non-Cartesian MRI quality can be improved by iterative reconstruction with constraints. Nearly all iterative MRI reconstructions rely on L2 least square error as a measure of data fidelity, which assumes that data are influenced by stationary Gaussian white noise. However, MR signals can be nonstationary due to noise correlation between multiple coils [1] and acoustic noise [2]. Thus, it may be advantageous to introduce L1 constrained least absolute deviation (L1-LAD) as an alternative measure of data fidelity, because LAD can be used for non-stationary non-Gaussian noise [3, 4]. From our previous Python non-uniform fast Fourier transform (pynufft) software [5, 6], we developed an L1 constrained LAD framework, which can carry out MR image reconstructions on a central processing unit (CPU) or a graphic processing unit (GPU) [6]. Methods and Implementation: L1-LAD can be described by the following L1-L1 scheme: u = argmin l ||y - Au||1 + k1 ||w1 (u)||1 + k2 ||w2(u)||1…(1). where u is the estimated image; l is the coefficient of LAD; A is the encoding matrix; y is the k-space data. w1, w2 are constraints which sparsify image u; k1, k2 are coefficients for constraints. Unlike the previous L2 least square error, here the data fidelity term is L1. To minimize the L1 norm, the variable-splitting method and a general shrinkage mapping were chosen [7]. The L1-LAD algorithm was incorporated into our pynufft software (previously developed for non-Cartesian MRI reconstruction). Pynufft was written in pure Python and integrated with GPU computing from version 3.2.9. It also provides identical APIs for CPU/GPU, which could enable rapid prototyping of new reconstruction algorithms on GPU. A high-resolution sagittal brain template [8] was used to test the L1-LAD algorithm in the software. Features illustrated at the exhibit: See Figure 1 for schematic diagram of the L1-LAD software. Figure 2 compares the L1-LAD and conventional density compensation method. L1-LAD exhibits a lower mean-square-error (MSE) than the conventional density compensation. Image details are clearly recovered with L1-LAD, while density compensation contains fine-grained artifacts in the image (see arrows).

123

S678

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 [5] Jyh-Miin Lin, Andrew J Patterson, Hing-Chiu Chang, Jonathan H Gillard, and Martin J Graves. An iterative reduced field-of-view reconstruction for periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) MRI. Medical Physics, 42(10):5757–5767, 2015. [6] Jyh-Miin Lin and Hsiao-Wen Chung. Pynufft: python non-uniform fast Fourier transform for MRI. In Building Bridges in Medical Sciences 2017, St John’s College, CB2 1TP Cambridge, UK, 2017. [7] Rick Chartrand. Shrinkage mappings and their induced penalty functions. In Acoustics, Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on, pages 1026–1029. IEEE, 2014. [8] Florent Lalys, Claire Haegelen, Jean-Christophe Ferre, Omar ElGanaoui, and Pierre Jannin. Construction and assessment of a 3-T MRI brain template. Neuroimage, 49(1):345–354, 2010.

726 An interactive, real-time, MRI simulator for smartphone/tablet S. Ma˚nsson Medical Radiation Physics, Lund University, Malmo¨/SWEDEN Purpose of the software: An MRI simulation tool was implemented as a smartphone/tablet app. The purpose was to create an intuitive tool, available to a broad range of healthcare professionals. The app is aimed at users who want to explore how scan parameters affect the MR image, with a higher degree of realism than existing smartphone apps, e.g. [1], but without the complexity of general-purpose simulators [2–4]. Methods and Implementation: Images of healthy volunteers were acquired with informed consent. The imaging protocol consisted of a single-slice, multi-echo turbo spin echo (TSE) sequence with three echo times, repeated with four repetition times. By manual selection of a point in any of the 12 TSE images, a signature for the corresponding tissue was created (Figure 1).

References: [1] Santiago Aja-Ferna´ndez and Antonio Trista´n-Vega. Influence of noise correlation in multiple-coil statistical models with sum of squares reconstruction. Magnetic Resonance in Medicine, 67(2):580–585, 2012. [2] Ziyue Wu, Yoon-Chul Kim, Michael CK Khoo, and Krishna S Nayak. Evaluation of an independent linear model for acoustic noise on a conventional MRI scanner and implications for acoustic noise reduction. Magnetic Resonance in Medicine, 71(4):1613–1620, 2014. [3] Jelena Markovic, Ruixun Zhang, and Lie Wang. Least absolute deviations method for sparse signal recovery. URL: https://math.mit. edu/research/undergraduate/spur/documents/2012Markovic.pdf, 2013. [4] Lie Wang. The penalized LAD estimator for high dimensional linear regression. Journal of Multivariate Analysis, 120:135–151, 2013.

123

After selecting N signatures, the following equation is solved for the vector C in each pixel: [S1 S2… SN]12xNCNx1 = Y12 9 1.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 where Sn denotes the n-th signature vector, C is a Nx1 amplitude vector, and Y represents the pixel values from the 12 TSE images. All TSE images can thus be represented by the signature vectors Sn multiplied by the N maps C (hereafter denoted clusters), as shown in Figure 2 for 1, 2 and 3 clusters.

Once the clusters C are computed, an arbitrary imaging sequence can be simulated by calculating new signatures based on the signal equation for the sequence and the relaxation times of the clusters. With this computation model, the signal equation needs to be evaluated once for each cluster, not for each pixel of the image. The simulation model was implemented on Apple iPhones 5, 6, 7, and an iPad Air. Sequence parameters could be changed interactively (Figure 3) and the simulated image was updated without noticable delay.

S679

727 FAst Nonlinear Susceptibility Inversion (FANSI Toolbox for QSM) C. Milovic1, B. Bilgic2, B. Zhao2, J. Acosta-Cabronero3, C. Tejos1 1 Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Santiago/CHILE, 2radiology, mgh, Boston/UNITED STATES OF AMERICA, 3Wellcome Trust Centre for Neuroimaging, University College London, London/UNITED KINGDOM Purpose of the software: Quantitative susceptibility mapping (QSM) inversion is often formulated as a minimization of a functional consisting of regularization and data fidelity terms. Data fidelity comprises a susceptibility-to-field relationship based on a phase noise distribution. Whereas such distribution can be approximated by a Gaussian function in most cases, notably it breaks down when SNR is too low1. To mitigate such shortcoming, a nonlinear data fidelity term that projects data into the complex image domain has been proposed2. Such formulation successfully reduces streaking artifacts, improves noise mitigation and returns more accurate susceptibility values3; however, it is computationally expensive, i.e. much slower than previously proposed QSM methods. The FANSI toolbox hereby presented contains a fast nonlinear solver4,5 that achieve computation speeds comparable to linear formulations, and is extended to work with both, total variation (TV) and total generalized variation (TGV) regularizations6-10. Methods and Implementation: FANSI utilizes the Alternating Direction Method of Multipliers (ADMM) to augment the cost functional with additional splitting variables. This formulation leads to a number of sub-problems, that effectively splits the regularization and data fidelity terms. As result, closed-form solutions are found to each sub-problem10. A Newton–Raphson iterative procedure solves the decoupled nonlinear sub-problem. FANSI is implemented as an open-source Matlab toolbox. Features illustrated at the exhibit: Reconstructions (using a range of quality scores, Figure 1) were comparable to nonlinear MEDI1,11,12 with more than an order of magnitude improvement in computational efficiency. Experiments showed a speed gain of up to 359 (Figures 2 and 3) using TV regularization over MEDI13. In addition, FANSI allows spatially adapted regularization and data fidelity terms for greater algorithmic control and flexibility.

Features illustrated at the exhibit: All features of the app will be illustrated at the exhibit, including real-time changes of sequence parameters and field strength, which will visually affect image contrast, SNR, chemical shift, fat–water in-phase/opposed phase and geometrical distortion. References: [1]. MRI Sim. http://www.icatsoftware.co.uk/radiology.htmls. [2]. Benoit-Cattin H, Collewet G, Belaroussi B, Saint-Jalmes H, Odet C. The SIMRI project: a versatile and interactive MRI simulator. Journal of magnetic resonance 2005;173(1):97–115. [3].Stocker T, Vahedipour K, Pflugfelder D, Shah NJ. High-performance computing MRI simulations. Magnetic resonance in medicine 2010;64(1):186–193. [4]. Xanthis CG, Venetis IE, Chalkias AV, Aletras AH. MRISIMUL: a GPU-based parallel approach to MRI simulations. IEEE transactions on medical imaging 2014;33(3):607–617.

123

S680

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 1. Gudbjartsson H, Patz S. The Rician distribution of noisy MRI data. Magnetic Resonance in Medicine. 1995;34(6):910–914. 2. Liu T, Wisnieff C, Lou M, Chen W, Spincemaille P, Wang Y. Nonlinear formulation of the magnetic field to source relationship for robust quantitative susceptibility mapping. Magn Reson Med. 2013;69:467–76. 3. Wang S, Liu T, Chen W, Spincemaille P, Wisnieff C, Tsiouris a J, Zhu W, Pan C, Zhao L, Wang Y. Noise Effects in Various Quantitative Susceptibility Mapping Methods. Biomedical Engineering, IEEE Transactions on. 2013;60:3441–3448. 4. Milovic C, Bilgic B, Zhao B, Acosta-Cabronero J, Tejos C. A Fast Algorithm for Nonlinear QSM Reconstruction with Variational Penalties. Proceedings of the Fourth International Workshop on MRI Phase Contrast & Quantitative Susceptibility Mapping. 2016;1:132. 5. Milovic C, Bilgic B, Zhao B, Acosta-Cabronero J, Tejos C. Fast Nonlinear Susceptibility Inversion (QSM) with Variational Regularizations. MRM. Submitted. 6. Bredies K, Kunisch K, Pock T. Total Generalized Variation. SIAM J. Imaging Sci. 2010;3:492–526. 7. Knoll F, Bredies K, Pock T, Stollberger R. Second order total generalized variation (TGV) for MRI. Magn Reson Med. 2011;65:480–491. 8. Yanez F, Fan A, Bilgic B, Milovic C, Adalsteinsson E, Irarrazaval P. Quantitative Susceptibility Map Reconstruction via a Total Generalized Variation Regularization. 2013 International Workshop on Pattern Recognition in Neuroimaging. 2013:203–206. 9. Langkammer C, Bredies K, Poser B a., Barth M, Reishofer G, Fan AP, Bilgic B, Fazekas F, Mainero C, Ropele S. Fast quantitative susceptibility mapping using 3D EPI and total generalized variation. NeuroImage. 2015;(2014). 10. Bilgic B., Chatnuntawech I., Langkammer C., Setsompop K.; Sparse Methods for Quantitative Susceptibility Mapping; Wavelets and Sparsity XVI, SPIE 2015. 11. Liu T, Liu J, De Rochefort L, Spincemaille P, Khalidov I, Ledoux JR, Wang Y. Morphology enabled dipole inversion (MEDI) from a single-angle acquisition: Comparison with COSMOS in human brain imaging. Magn Reson Med. 2011;66:777–783. 12. Liu J, Liu T, De Rochefort L, Ledoux J, Khalidov I, Chen W, Tsiouris AJ, Wisnieff C, Spincemaille P, Prince MR, et al. Morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map. NeuroImage. 2012;59:2560–2568. 13. Milovic C, Bilgic B, Zhao B, Acosta-Cabronero J, Tejos C. A Fast Algorithm for Nonlinear QSM Reconstruction. Proceedings of the ISMRM17, Hawaii, USA. Eposter 3669.

728 GPU enabled implementation of 3 Compartment Leaky Tracer Kinetic Model (LTKM) for DCE-MRI D. Rathore, R. Rathore Imaging R&D, ADISL, Kanpur/INDIA

Acknowledgments: ACT1416 Programa PIA CONICYT, FONDECYT 1161448, Becas Doctorado Nacional F:21150369. References:

123

Purpose of the software: The frequent breakthroughs and a worldwide interest in R&D in MRI (and related protocols viz. DCE-MRI, MRS, DTI among others) have generated a need for tools and frameworks[2,3,4] that allow rapid development of research grade software for fast processing of data. We are presenting NVIDIA GPU based implementation of the 3 compartment Leaky Tracer Kinetic Model (LTKM) [1,2,6] for Analysis of DCE-MRI data. NVIDIA CUDA Library was used in the C++ based implementation.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 GPU based Matrix computations are faster than the standard CPU based one due to the higher number of computing cores available to perform parallel processing. C++ based implementation means that the memory restrictions associated with Virtual Machine based platforms do not arise here. The 3 Compartments in the LTKM Intravascular Plasma Compartment Intravascular Plasma Space Space 1 Compartment The permeable space (region of 2 bidirectional exchange) Extracellular Leakage space (a space of Extravascular Space Compartment unidirectional flow from which the (EES) 3 contrast does not flow back into the vasculature) If the leakage compartment is absent, LTKM behaves like GTKM [Tofts], and if the permeability compartment is absent LTKM behaves like Patlak model[Patlak] or the tissue uptake model.

Methods and Implementation: Methodology and results of DCEMRI analysis using LTKM have been validated in referenced publications[1]. The NVIDIA GPU based tool [Figure 1, 2] is supported on a PC with Microsoft Windows and this plugin was developed using Microsoft Visual Studio. GPU computing is done using NVIDIA CUDATM libraries [7].

S681 § Visualization—2D data is visualized in runtime or on need-basis. 3D Visualization is being done using VTK[15]. References: 1. Sahu et al., 2013 Sep, 38(3):677–88, JMRI. 2. Rathore RKS et al., 2012, ISMRM 20, 1976. 3. Rathore D et al., 2012, ESMRMB 29, 844. 4. Rathore D et al., 2011, ESMRMB 28, 761. 5. Rathore DK et al., 2011, ISMRM 19, 4891. 6. Rathore RKS et al., 2011, ISMRM 19, 2053. 7. http://en.wikipedia.org/wiki/CUDA. 8. Tofts PS, et al., 1999, JMRI, 10:223–32. 9. Sourbron SP et al., Phys Med Biol 2012. 10. Li KL et al., Br J Radiol 2003;76:39–50. 11. Patlak CS et al., J Cereb Blood Flow Metab 1983;3:1–7. 12. Rishi Awasthi et al., 2010, JCAT. Jan, 34(1):82–8. 13. Purwar A et al., 2006, ISMRM 14. 14. Singh A et al., 2007, JMRI, 26:871–880. 15. Singh A et al., 2007, ISMRM-ESMRMB (2240). 16. Rathore RKS et al., 2008, ISMRM, 1884. 17. http://www.vtk.org. 18. http://www.gpu.io.

729 Simple and easy-to-use graphical user interface (GUI) for accurate and stable mapping of T2 relaxation values D. Radunsky, N. Ben-Eliezer Department of Biomedical Engineering, Tel Aviv University, Tel Aviv/ ISRAEL

Features illustrated at the exhibit: We illustrate post-processing of DCE-MRI data over a mix of CPU and GPU using NVIDIA CUDATM libraries[3]: § Descalping[13]—removes non-brain anatomies. § Pre-contrast tissue parameter estimation and quantitation of absolute tracer concentration from measured signal intensity[14]. § A piecewise linear model for automatic estimation of BAT and other segmentation parameters like slope of line segments[15]. § Automatic measurement of AIF along with the option of PVE correction for individual patients[16]. § Both, tracer kinetic/generalized tracer kinetic and first pass analysis models with correction of leakage effect on CBV estimation are implemented[14,16,8]. § ROI analysis module for DCE-MRI quantitation.

Purpose of the software: Target audience: physicians and biologists using T2 contrast, or interested in quantitative mapping of T2. T2 contrast is used for a wide range of non-invasive diagnosis and characterization of pathologies. Traditionally, MRI produces qualitative image-contrast, which varies between scanners, scan-settings, and moreover, is subjected to reader bias. Quantitative evaluation of T2, on the other hand, offers a more accurate characterization of the tissue and has been proven valuable for a variety of applications125. Reliable assessment of T2, however, is challenging due to the inherent deviation of multi spin-echo (MSE) protocols from the theoretical exponential model S(t) = S0exp[–t/T2] as a result of stimulated and indirect echoes, non-rectangular slice profiles, and inhomogeneous transmit (B1 1 ) field profiles. Several methods were developed to overcome these effects6–11. Our software is based on the Echo Modulation Curve (EMC) algorithm12,13, which has been shown to deliver the true T2 value of the tissue—closely matching values obtained using reference single spinecho acquisitions14. We present software for producing quantitative T2 maps based on rapid MSE data. The software is accessed via an intuitive and easy-to-use GUI (Figure 1) providing an automatic T2 fitting and allowing users to probe T2 and proton density (PD) values in selected regions of interest (ROIs). Methods and Implementation: Implementation: The EMC algorithm and GUI were programmed in MATLAB and C??. Algorithm is based on matching experimental data to a database of simulated decay curves, matching the specific protocol used for acquisition12. Input: MSE data in raw k-space or in DICOM format. Output: T2 and PD maps. Presented GUI (Figure 1) also enables to automatically generate tailored databases. Stability benchmark: Data were collected for single subject using six different MSE parameter sets (Table 1), and for a group of 37 healthy volunteers using a single parameter set. T2 maps were reconstructed using the EMC algorithm12 and using conventional mono-exponential

123

S682

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

model. Mean, standard deviation (SD), and coefficient of variation (CV) were calculated for 5 brain ROIs. Stability was evaluated based on three categories (Table 2): (1) Global inter scan variability; (2) Specific inter- and intra- scanner variability; (3) Inter subject variability across 37 volunteers.

References: [1] Siemonsen S et al. Stroke 2009; 40(5): 1612–16. [2] Lund H et al. Acta Neurol Scand 2012; 125(5): 338–44. [3] Eitel I, Friedrich MG. T2. J Cardiovasc Magn Reson 2011; 13(1): 13. [4] Liu W et al. Magn Reson Med 2011; v. 65(5), p. 1400–6. [5] Pan J et al. Radiology 2011; v. 261(2), p. 507–15. [6] Warntjes J et al. Magn Reson Med 2007; v. 57(3), p. 528–37. [7] Lukzen N et al. J Magn Reson 2009; v. 196(2), p. 164–9. [8] Lebel R, Wilman A. Magn Reson Med 2010; v. 64(4), p. 1005–14. [9] Doneva M et al. Magn Res Med 2010; v. 64(4), p. 1114–20. [10] Prasloski T et al. Magn Reson Med 2012; v. 67(6), p. 1803–14. [11] Ma D et al. Nature 2013; v. 14(495(7440)), p. 187–92. [12] Ben-Eliezer N et al. Magn Reson Med 2015; 73(2): 809–17. [13] Ben-Eliezer N et al. Magn Reson Med 2016; 75(3): 1346–54. [14] McPhee KC, Wilman AH. Magn Reson Med. 2016; doi: 10.1002/mrm.26285. Financial support: NIH Grants: P41 EB017183; RO1 EB000447.

730 Pulseq: A rapid and hardware-independent pulse sequence prototyping framework

Features illustrated at the exhibit: Software GUI designed for clinicians and basic scientists. Features include (1) Load raw k-space or DICOM data; (2) Construction of T2 and PD maps; (3) Generation of synthetic T2 weighted images; (4) Arbitrary ROI statistics; (5) Exporting maps in DICOM, NIFTI, or MATLAB format; (6) Automatic generation of new EMC databases.

123

S. Kroboth1, K. Layton2, F. Jia1, S. Littin1, H. Yu1, J. Leupold1, J.-F. Nielsen3, T. Stoecker4, M. Zaitsev1 1 University Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Dept. of Radiology, Medical Physics, Freiburg/GERMANY, 2University of South Australia, Institute for Telecommunications Research, Mawson Lakes/ AUSTRALIA, 3Department of Biomedical Engineering, University of Michigan, Michigan/UNITED STATES OF AMERICA, 4DZNE, German Center for Neurodegenerative Diseases, Bonn/GERMANY Purpose of the software: Implementing MR sequences often involves extensive programming on vendor-specific platforms, which can be time-consuming and costly. Even more so if research sequences need to be implemented on several platforms simultaneously, for example at different field strengths or on machines of different vendors. We introduce an alternative programming environment called Pulseq [1] that is hardware-independent, open-source, and promotes rapid sequence prototyping. Central to this approach is a novel file format which allows for the efficient description of hardware events and timing information required for an MR pulse sequence. Platform-dependent interpreter modules convert the file to appropriate instructions to run the sequence on MR hardware.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Sequences are designed in high-level programming languages such as MATLAB, Python or with graphical user interfaces (i.e. JEMRIS [2] and GPI Lab [3]). This concept allows for the decoupling of sequence design (hardware-independent) and sequence execution (hardwaredependent). Furthermore, when using JEMRIS, sequences can be both executed on real hardware and simulated, allowing for comparison of real and virtual experiments. Pulseq is an excellent tool for teaching MR physics and has the potential to scale well to multi-centre studies. The source code can be downloaded from Github [4]. Methods and Implementation: The main components of the Pulseq environment are illustrated in Fig. 1. The high-level sequence can be described directly in MATLAB or Python using functions from a provided toolbox [4]. Alternatively, sequences can be programmed using the graphical interface of the JEMRIS simulation packages [2] or GPI Lab [3]. Regardless of the choice of high-level interface, a sequence file is created containing low-level sequence instructions such as RF pulses, gradients, ADC events and delays. This file can then be executed on various platforms via hardware-dependent interpreter modules.

Figure 1 also shows the results of a gradient echo sequence executed on three different hardware platforms: a 3T Siemens Trio equipped with a single-channel wrist RF coil (Siemens Healthcare, Erlangen, Germany); a 3 T GE Discovery MR750 with a 8 channel head coil (GE Healthcare, Waukesha, WI, USA); and a 9.4 T Bruker BioSpec MRI with a single-channel rat coil (Bruker Biospin, Ettlingen, Germany). Features illustrated at the exhibit: We will show how basic sequences can be implemented in Pulseq using several of the mentioned design tools. References: [1] Layton KJ, Kroboth S, Jia F, Littin S, Yu H, Leupold J, Nielsen JF, Sto¨cker T, Zaitsev M. Pulseq: A rapid and hardware-independent pulse sequence prototyping framework. Magn Reson Med 2017; 77:1544–52. [2] Sto¨cker T, Vahedipour K, Pflugfelder D, Shah NJ. High-performance computing MRI simulations. Magn. Reson. Med. 2010;64:186–193. [3] Keerthi Sravan R, Shaik I A, Kroboth S, Zaitsev M, and Geethanath S. Implementation of Pulseq in GPI Lab. Proc. ISMRM 2017. [4] http://pulseq.github.io/.

S683

731 ConsTru: an optimal B0 shimming solution P. Chang1, S. Nassirpour1, A. Henning2 1 Ultra High Field for MRI, Max Planck Institute for Biological Cybernetics, Tuebingen/GERMANY, 2Institute of Physics, ErnstMoritz-Arndt University Greifswald, Greifswald/GERMANY Purpose of the software: B0 shimming is a vital pre-scan adjustment that directly affects the quality of the data. However, vendor implemented B0 shimming can be time-consuming and sub-optimal. For smaller regions-of-interest (ROI), such as single voxel spectroscopy, the B0 shim can be especially poor and exceed the hardware constraints. We present a software application that can calculate the optimal B0 shims for ROIs of any size, for arbitrary ROIs and also single-voxels. The process is fast, non-iterative and the shim values are always within the hardware constraints. The software package can easily be installed and used on any MR and shim system, and has the ability to easily be calibrated to account for any imperfections of the system. Methods and Implementation: The B0 shim values are calculated from B0 maps using an optimized and robust algorithm. We use the ConsTru algorithm1 that has previously been shown to give the optimal shim values for any application: whole brain, single slice, single voxel. The algorithm considers the hardware constraints of the system and regularizes the problem accordingly so that the shim values are the best for the given shim system. The software uses B0 maps in DICOM format. The region of interest can be defined by the user or copied from the scan protocol and the B0 shim values are then calculated. Figure 1 shows the superiority of the ConsTru shimming algorithm over a variety of other B0 shimming algorithms currently used in the field in the case of a challenging shimming application (single voxel located in the pre-frontal cortex at 7 T). The resulting shimmed B0 maps and the standard deviation of the frequency shifts are shown for each algorithm. Figure 2 shows the resulting shim quality using ConsTru at 7 and 9.4 T over 66 volunteers for different ROIs. The software can also be modified to account for non-ideal spherical harmonic fields. Real fields models and reference fields can be used to calculate the optimal shim values2.

123

S684

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

References: [1] Nassirpour S, et al. MRM 2017. [2] Chang P, et al. MRM 2017.

732 ARIANNA: a web-based platform for collaborative research in neuroimaging of Autism Spectrum Disorders A. Giuliano1, A. Retico1, P. Bosco1, S. Arezzini1, S. Calderoni2, A. Ciampa1, S. Coscetti1, S. Cuomo3, L. De Santis4, D. Fabiani1, M.E. Fantacci5, E. Mazzoni1, P. Mercatali3, G. Miscali4, M. Pardini4, M. Prosperi2, M. Prosperi2, F. Romano3, E. Tamburini6, M. Tosetti7, F. Muratori2 1 Pisa, National Institute for Nuclear Physics (INFN), Pisa/ITALY, 2 Developmental Neuroscience, IRCCS Stella Maris Foundation, Pisa/ ITALY, 3Institute of Legal Information Theory and Techniques (ITTIG), National Research Council (CNR), Florence/ITALY, 4S.r.l., NET7, Pisa/ITALY, 5Physics Department, University of Pisa, Pisa/ ITALY, 6S.r.l., I?, Florence/ITALY, 7Imago7 Foundation, Imago7 Foundation, Pisa/ITALY

Features illustrated at the exhibit: The software is a simple, standalone executable. It can run on Windows 7 and upwards (and Linux binaries can also be compiled). There is a single-voxel version and an arbitrary ROI version available. The arbitrary ROI version also comes with an autosegmentation feature. The two versions are shown in Figure 3. After the ROI is defined, the shim values are calculated immediately and no further iterations are required. The shim values as well as the defined ROIs can be saved to file.

123

Purpose of the software: The complexity of Autism Spectrum Disorders (ASD) requires the implementation of sophisticated algorithms to infer disease diagnosis and to get the most from the many heterogeneous variables that describe affected individuals. Despite the structural and functional brain patterns highlighted in subjects with ASD through magnetic resonance imaging (MRI) studies, a number of open issues still remain1–3. The ARIANNA project has developed a publicly available web-based platform for neuroimaging data analysis accessible to the entire community of ASD researchers (). The main goals of the ARIANNA research environment are: to analyze neuroimaging data acquired in multiple sites with multivariate approaches based on machine learning; to identify structural and functional brain alterations that can support the diagnosis of ASD; to detect neuroimaging-based criteria for the stratification of the population with ASD to develop personalized treatments. Methods and Implementation: The web-based architecture, the computing infrastructure and the collaborative analysis workflows at the basis of ARIANNA interdisciplinary working environment are presented, and the full functionality of the research platform is demonstrated. The ARIANNA infrastructure is hosted at the INFN Pisa scientific computing center, where dedicated storage and computing units have been designed to satisfy the neuroimaging data analysis requirements. Secure data handling and storage are guaranteed, as well as a fast access to computational resources. To test the efficiency of the available resources, we run the FreeSurfer recon-all

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S685

tool both on a proprietary data sample of structural MRI (sMRI) of 152 subjects and on the public sample of 1112 subjects of the Autism Brain Imaging Data Exchange (ABIDE) project (. Features illustrated at the exhibit: The data entry system allows the insertion of demographic and clinical variables characterizing each subject, such as gender, age at MRI, motivation for the exam, IQ level, and many clinical variables related to ASD condition (Fig. 1). The platform accepts the upload of the sMRI, the Diffusion Tensor Imaging and the resting-state functional MRI scans. One or more of these acquisitions can be uploaded and they can be inserted at different time points. The ARIANNA database can be easily interrogated through a dedicated web interface (Fig. 2). A selection of the available subjects can be made according to criteria such as gender, ranges of age, IQ level etc. Therefore, the outcomes of the query are a list of subjects along with their variables and the location of the corresponding neuroimaging files which can be redirected to the desired workflows. Some preliminary results obtained within the ARIANNA framework are presented.

123

S686

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Rossendorf, Dresden/GERMANY, 3Diagnostic Physics, Oslo University Hospital, Oslo/NORWAY, 4Radiology and Nuclear Medicine, Academic Medical Center, Amsterdam/NETHERLANDS, 5 Sunnybrook Research Institute, University of Toronto, Toronto/ CANADA, 6Institute of Psychiatry, King’s College London, London/ UNITED KINGDOM, 7Radiology, University Medical Center, Utrecht/NETHERLANDS, 8Neurology, Radboud University Medical Center, Nijmegen/NETHERLANDS, 9Radiology, Academic Medical Center, Amsterdam/NETHERLANDS, 10C.J. Gorter Center for high field MRI, department of Radiology, LUMC, Leiden/NETHERLANDS, 11 Brain Repair & Rehabilitation, Institute of Neurology, University College London, UK, London/UNITED KINGDOM, 12Department of Radiology, Academic Medical Center Amsterdam, Amsterdam/ NETHERLANDS, 13Kate Gleason College of Engineering, Rochester Institute of Technology, New York/UNITED STATES OF AMERICA

References: 1. Ecker C, Murphy D. Neuroimaging in autism–from basic science to translational research. Nat Rev Neurol 2014;10:82–91. doi: 10.1038/nrneurol.2013.276. 2. Retico A, Tosetti M, Muratori F, Calderoni S. Neuroimaging-based methods for autism identification: A possible translational application? Funct Neurol 2014;29:231–9. 3. Volkmar FR, State M, Klin A. Autism and autism spectrum disorders: diagnostic issues for the coming decade. J Child Psychol Psychiatry 2009;50:108–15. doi:10.1111/j.1469-7610.2008.02010.x.

Purpose of the software: Arterial spin labeling (ASL) perfusion MRI1 is rapidly maturing as a promising biomarker2 and pharmacological monitoring agent3. To advance this goal, there is a growing need for standardization of ASL image processing and quality control (QC). Here, we present ExploreASL, a non-commercial software package designed to harmonize image processing for ASL perfusion images for single- and multi-center studies4, 5. In addition, ExploreASL combines key structural image processing tools from recent literature to differentiate perfusion from structural effects. Initiated through the EU-funded COST-action ‘‘ASL In Dementia’’6, ExploreASL is a collaborative framework that facilitates cross-pollination between image processing method developers and clinical investigators. It has currently processed * 4000 ASL images from all major MRI vendors and ASL sequences, and a variety of patient populations, from * 30 studies. The ultimate goal is to combine data from multiple studies to identify common and differential perfusion patterns. This may enhance knowledge of the interplay of perfusion and structural changes in neurodegenerative pathophysiology. Methods and Implementation: ExploreASL features: easy data import from various image formats7, easily configurable, fully automatic, progress tracking and error logging—can be paused at any stage, supports all major vendors and ASL sequences4, 5, 8–11, multiOS and parallel processing support, optimized segmentation and registration with single interpolation from raw data to 1.5 9 1.5 9 1.5 mm common space12, 13, ATT estimation from single-PLD ASL14, acquisition artifacts filter, semi-automatic scan QC for subject exclusion14, site bias-field creation to remove site effects, population data-driven analysis mask. ExploreASL is written in Matlab (MathWorks, MA, USA) and is based on SPM12 (University College London, UK).

733 ExploreASL: image processing toolbox for multi-center arterial spin labeling population analyses H.J. Mutsaerts1, J. Petr2, E. Lysvik3, A. Schrantee4, Z. Shirzadi5, F. Zelaya6, I. Groote3, O. O’Daly6, J. Kuijer1, J. De Bresser7, E. Richard8, M.W.A. Caan9, M. Van Osch10, X. Golay11, L. Reneman4, B. Macintosh5, M. Masellis5, J. Hendrikse7, F. Barkhof1, A. Bjornerud3, A.J. Nederveen12, I. Asllani13, P. Groot4 1 Radiology, VU University Medical Center, Amsterdam/ NETHERLANDS, 2PET Center, Helmholtz-Zentrum Dresden-

123

Features illustrated at the exhibit: ExploreASL allows for a quick data-to-paper approach, with QC and results from each processing step calculated and saved in individual (Figure 2a), and population-

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 wide (Figure 3b) NIfTI and JPG images and Excel tables containing results and group statistics and comparisons for large and small anatomical and vascular ROIs 15-18. ExploreASL incorporates a detailed manual, with screenshots to get quickly acquainted with its many features. ExploreASL can be downloaded from GitHub upon request, including regular updates. Contact [email protected] for more information.

References: 1. Alsop DC, Detre JA, Golay X, et al. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 2015; 73(1):102–116. 2. Steketee RM, Bron EE, Meijboom R, et al. Early-stage differentiation between presenile Alzheimer’s disease and frontotemporal dementia using arterial spin labeling MRI. Eur Radiol 2015. 3. Wang DJ, Chen Y, Fernandez-Seara MA, et al. Potentials and challenges for arterial spin labeling in pharmacological magnetic resonance imaging. J Pharmacol Exp Ther 2011; 337(2):359–366. 4. Mutsaerts HJMM, Petr J, Thomas DL, et al. Comparison of arterial spin labeling registration strategies in the multi-center GENetic frontotemporal dementia initiative (GENFI). J Magn Reson Imaging 2017. 5. Mutsaerts HJ, van Osch MJ, Zelaya FO, et al. Multi-vendor reliability of arterial spin labeling perfusion MRI using a near-identical sequence: implications for multi-center studies. Neuroimage 2015; 113:143–152. 6. http://www.aslindementia.org/. In: 2017. 7. Gorgolewski KJ, Auer T, Calhoun VD, et al. The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Sci Data 2016; 3:160044. 8. Mutsaerts HJ, Steketee RM, Heijtel DF, et al. Inter-vendor reproducibility of pseudo-continuous arterial spin labeling at 3 T. PLoS One 2014; 9(8):e104108. 9. Mutsaerts HJ, Steketee RM, Heijtel DF, et al. Reproducibility of pharmacological ASL using sequences from different vendors: implications for multicenter drug studies. MAGMA 2015. 10. Steketee RM, Mutsaerts HJ, Bron EE, et al. Quantitative Functional Arterial Spin Labeling (fASL) MRI–Sensitivity and Reproducibility of Regional CBF Changes Using Pseudo-Continuous ASL Product Sequences. PLoS One 2015; 10(7):e0132929. 11. Heijtel DF, Mutsaerts HJ, Bakker E, et al. Accuracy and precision of pseudo-continuous arterial spin labeling perfusion during baseline and hypercapnia: a head-to-head comparison with (1)(5)O H(2)O positron emission tomography. Neuroimage 2014; 92:182–192. 12. Ashburner J, Friston KJ. Nonlinear spatial normalization using basis functions. Hum Brain Mapp 1999; 7(4):254–266. 13. Ashburner J. A fast diffeomorphic image registration algorithm. Neuroimage 2007; 38(1):95–113. 14. Mutsaerts HJ, Petr J, Vaclavu L, et al. The spatial coefficient of variation in arterial spin labeling cerebral blood flow images. J Cereb Blood Flow Metab 2017:271678X16683690.

S687 15. Desikan RS, Segonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage 2006; 31(3):968–980. 16. Mazziotta J, Toga A, Evans A, et al. A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM). Philos Trans R Soc Lond B Biol Sci 2001; 356(1412):1293–1322. 17. Mutsaerts HJ, van Dalen JW, Heijtel DF, et al. Cerebral Perfusion Measurements in Elderly with Hypertension Using Arterial Spin Labeling. PLoS One 2015; 10(8):e0133717. 18. Tatu L, Moulin T, Bogousslavsky J, et al. Arterial territories of the human brain: cerebral hemispheres. Neurology 1998; 50(6):1699–1708. 19. Schmidt P, Gaser C, Arsic M, et al. An automated tool for detection of FLAIR-hyperintense white-matter lesions in Multiple Sclerosis. Neuroimage 2012; 59(4):3774–3783. 20. Ashburner J, Ridgway GR. Symmetric diffeomorphic modeling of longitudinal structural MRI. Front Neurosci 2012; 6:197. 21. Shirzadi Z, Crane DE, Robertson AD, et al. Automated removal of spurious intermediate cerebral blood flow volumes improves image quality among older patients: A clinical arterial spin labeling investigation. J Magn Reson Imaging 2015; 42(5):1377–1385. 22. Beaumont H. Multimodal Magnetic Resonance Imaging of Frontotemporal Lobar Degeneration. In: University of Manchester, Faculty of Medical and Human Sciences; 2015. pp. 102–126. 23. Asllani I, Borogovac A, Brown TR. Regression algorithm correcting for partial volume effects in arterial spin labeling MRI. Magn Reson Med 2008; 60(6):1362–1371. 24. Mutsaerts HJ, Richard E, Heijtel DF, et al. Gray matter contamination in arterial spin labeling white matter perfusion measurements in patients with dementia. Neuroimage Clin 2013; 4:139–144. 25. Petr J, Mutsaerts H, De Vita E, et al. Deformation and resolution issues in partial volume correction of 2D and 3D arterial spin labeling data. In: International Society of Magnetic Resonance in Medicine; 2016.

734 OpenMRSLab: An open-source software repository for Magnetic Resonance Spectroscopy data analysis tools B. Rowland Cardiff University Brain Research Imaging Centre, Cardiff University, Cardiff/UNITED KINGDOM Purpose of the software: The ISMRM MRS consensus group recently published a statement paper calling for greater standardisation of MRS analysis methods [1]. The widespread use of homegrown tools makes it difficult to reproduce results or even to implement described methods, as do closed source tools such as jMRUI and LCModel. While there are a number of open-source packages for working with MRS data including CIVIC, TARQUIN and Vespa, these all focus on providing a GUI application to apply a largely fixed-function processing pipeline with little flexibility. This works well for standard analyses, but it is more problematic when a project needs something non-standard. OpenMRSLab is an open-source project designed to offer an alternative approach by providing an accessible interface to the processing code itself, making it easy to flexibly combine pre-existing standard components with project specific custom steps into a complete processing pipeline, then to share that with other researchers and labs. Methods and Implementation: The core of the OpenMRSLab project is the Suspect package. Written entirely in Python, Suspect

123

S688 supports loading data from many different file formats, including DICOM and the major vendor private formats. Going beyond the basic processing steps such as apodisation and zero filling, the library also includes more sophisticated algorithms including SVD-based channel combination, spectral registration for frequency correction and HSVD water suppression. For metabolite quantification Suspect is fully integrated with both LCModel and TARQUIN for basis set fitting. It also provides an AMARES style singlet fitting routine intended primarily for X-nuclei quantification. Suspect works particularly well with the Jupyter notebook, a browser based interactive Python session which combines code, plots and notes.

OpenMRSLab can be easily installed using Docker. The project provides a downloadable container which has Suspect, TARQUIN, Jupyter and several other useful packages pre-installed and ready-touse.

Features illustrated at the exhibit: All components of the standard MRS processing pipeline will be presented, including channel combination, frequency correction, water suppression and quantification. In addition co-registration with imaging data and visualisation will be demonstrated. There is also the possibility to show X-nuclei processing using 31P data. References: [1] Oz, et al. (2014) Clinical proton MR spectroscopy in central nervous system disorders. Radiology. 270(3):658–79.

735 Medical imaging in the browser with the A* Medical Imaging (AMI) toolkit N. Rannou1, J.L. Bernal-Rusiel2, D. Haehn3, P.E. Grant2, R. Pienaar2

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Medical, Eunate Technology S.L., Sopela/SPAIN, 2Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Childrens Hospital, Boston/UNITED STATES OF AMERICA, 3Computer Science, Harvard University, Cambridge/UNITED STATES OF AMERICA 1

Purpose of the software: We present AMI (A* Medical Imaging) [1], a new JavaScript toolkit for medical imaging on the web. Popular currently available libraries are often very specialized and may lack some core imaging features: some are 2D only with no support for popular 3D models [2–4], others might support concurrent 2D and 3D visualization but often these are not in the same rendering window [5, 6], many have limited UI tool elements [7, 8], GPU acceleration is either limited or not present in most existing solutions [3–8], etc. AMI is a single toolkit that includes 2D and 3D visualization of most popular medical imaging data formats, the ability to render 2D/3D images and models concurrently in the same scene, allows for real time interaction and also provides a set of useful UI elements (such as rulers, painters, selectors, etc.). Methods and Implementation: The framework is written in JavaScript and is developed through continuous integration [9] for quality assurance. It has a modular architecture that makes it easily extensible, testable and understandable. WebGL support is provided by layering onto three.js [10]. A flexible design means that actual image and model data parsing is not part of the core library but external parsers can be easily linked to the library. Data parsers: DICOM [11], NIFTI [12], NRRD [13], MHD/RAW [1]. Model parsers: VTK [10], STL [10], TRK [1]. UI tools are designed to work seamlessly in 2D and 3D. Existing UI elements can also be simply combined to create new elements. Tools available (2D and 3D): distance measurements, angle measurements, annotations. AMI is open source to promote transparency, collaboration and improve code quality. New features submission and bug fixes are encouraged. All contributions are carefully reviewed and tested before integration to the main repository. Features illustrated at the exhibit: All AMI features through live examples publicly available online.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S689 [4] Bernal-Rusiel J.L., Rannou N., Gollub R., Pieper S., Murphy S., Robertson R., Grant P.E. and Pienaar R. Reusable Client-Side JavaScript Modules for Immersive Web-based Real-time Collaborative Neuroimage Visualization. Frontiers in Neuroinformatics, 2017. [5] Sherif T., Kassis N., Rousseau M.E., Adalat R., Evans A.C. BrainBrowser: distributed, web-based neurological data visualization. Frontiers in Neuroinformatics 2015. [6] Papaya—https://github.com/rii-mango/Papaya. [7] Haehn D., Rannou N., Ahtam B., Grant P.E. and Pienaar R. Neuroimaging in the Browser using the X Toolkit. Frontiers in Neuroinformatics, 2014. [8] Haehn D., Rannou N., Grant P.E. and Pienaar R. Slice:Drop. IEEE VisWeek, SciVis Poster Session, 2012. [9] Continuous Integration—https://en.wikipedia.org/wiki/Continuous_ integration. [10] three.js—https://github.com/mrdoob/three.js/. [11] dicomParser—https://github.com/chafey/dicomParser. [12] NIFTI Reader JS—Github—https://github.com/rii-mango/NIFTIReader-JS. [13] NRRD JS—Github—https://github.com/scijs/nrrd-js.

736 Implementation of Camino Diffusion MRI Toolkit: reconstruction of White Matter tissue in presence of demyelinating disease S. Oliviero, C. Del Gratta Neuroscience, Imaging and Clinical Sciences/ITAB, University of Chieti-Pescara G. D’Annunzio, Chieti/ITALY Purpose of the software: The software is an implementation of Camino Diffusion MRI Toolkit1 with the purpose of characterizing White Matter (WM) tissues with different degrees of demyelination and axonal loss. Methods and Implementation: Camino is an MRI toolkit that is capable to simulate the MR signal coming from a portion of white matter tissue (substrate) subjected to a magnetic field by performing a Monte Carlo diffusion simulation inside the substrate. The substrate consists of a free space containing cylinders representing the axons. A user has the chance to set the number of the axons per substrate and select one of the proposed configurations for the axons, i.e. parallel cylinders with equal or gamma-distributed radii and crossing cylinders. The first innovation of this implementation is the introduction of another type of substrate in which the parallel cylinders with gammadistributed radii, are covered by myelin (Figure 1).

References: Figure 1: Screenshot demonstrating concurrent 2D/3D visualization of data and models. Top left: 3D visualization of Axial, Coronal and Sagittal slices from a DICOM dataset, along with 2 STL models. Top right: 2D visualization of Axial slice with STL models overlay. Bottom left: 2D visualization of Sagittal slice with STL models overlay. Bottom right: 2D visualization of Coronal slice with STL models overlay. Figure 2: Screenshot demonstrating ray marching volume rendering. Figure 3: Screenshot demonstrating distance measurements tools. [1] AMI—https://github.com/FNNDSC/ami. [2] Cornerstone—https://github.com/chafey/cornerstone. [3] OHIF Viewers—https://github.com/OHIF/Viewers.

In the novel substrate, each axon is represented by a pair of coaxial cylinders: the inner one represents the axon, whilst the space between the inner and the outer cylinders is the myelin. In this work, we decided to characterize the intra-axonal space by including the effect of various organelles and other elements which are present in a real tissue. This resulted in the usage of the intraaxonal diffusion coefficient pointed by Stanisz3. The intra-myelin space is characterized by a pair of radial and tangential diffusion coefficients, as proposed by Trevor et al.4.

123

S690 Axon is separated from myelin by a permeable membrane (Stainisz3), while myelin has no membrane separating it from the surrounding space. Another peculiarity of the implementation is the possibility to simulate the processes of demyelination and axonal loss for the proposed substrate. A user can simulate a certain degree of demyelination by reducing the value of the g-ratio with respect to the healthy conditions. For the axonal loss, the software simulates the process by implementing the experimental observation5 of the selective death of cylinders with radius \ 1 lm. When a user set a given percentage p of dead axons with respect to the healthy condition of N axons, the software at first, generates a substrate containing N axons with gamma distributed radii, then it eliminates Np cylinders by extracting their radius from a sigmoidal distribution function, with the inflection point at 1 lm. Features illustrated at the exhibit:

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692 Calculation of rCBF, rCBV, and Tmax is performed using Deconvolutiion including Tikhonov Regularization. 2. Segmentation of the infarct-core and the penumbra The segmentation follows our approach [3] of an automated comparison of the left and right hemispheres. A potential lesion is determined based on local histogram comparisons, and inside this region, segmentations of the penumbra and the infarct core are generated by thresholding with generally accepted values Tmax [ 6 s and ADC \ 600 mm2/s [1]. 3. Calculation of an ASPECT—Score as a marker for affected perfusion-territories A quantitative score similar to ASPECTS [4] is calculated by atlasbased calculation of the percentage of the affected volume in several perfusion regions, and assigning a score from 10 to 0. 4. Generation of a report sheet All segmentations, time series, histograms and derived parameters are shown in the following comprehensive summary report sheet:

– Demyelination process – Axonal loss process References: 1 P. A. Cook et al. Camino: Open-Source Diffusion-MRI Reconstruction and Processing, 14th Scientific Meeting of the International Society for Magnetic Resonance in Medicine, Seattle, WA, USA, p. 2759, May 2006. 2 https://haleytitusmitchell.wordpress.com/research/research-projects/. 3 G. Reg. J. Stainisz. Diffusion MR in Biological Systems: Tissue Compartments and Exchange. Israel Journal of Chemistry 43: 33–44 (2003). 4 J. Trevor et al. Diffusion of Myelin Water. Magnetic Resonance in Medicine 56:381–385 (2006). 5 G. C. DeLuca et al. The contribution of demyelination to axonal loss in multiple sclerosis. Brain 129: 1507–1516 (2006). doi: 10.1093/brain/awl074.

737 KStroke: A computer-aided-diagnosis software package for fully automated processing, segmentation and analysis of diffusion/perfusion-MRI in acute stroke E. Kellner1, M. Reisert1, V.G. Kiselev1, H. Urbach2, K. Egger3 1 Dept. of Radiology, Medical Physics, Faculty of Medicine, Medical Center University of Freiburg, Freiburg/GERMANY, 2Dept. of Neroradiology, Faculty of Medicine, Medical Center University of Freiburg, Freiburg/GERMANY, 3Dept. of Neroradiology, Physics, Faculty of Medicine, Medical Center University of Freiburg, Freiburg/GERMANY Purpose of the software: We developed a fully automated software package for data processing of acute Stroke MRI data (Dynamicsusceptibility-contrast perfusion imaging, DSC-PWI and Diffusionweighted-Imaging, DWI). The processing pipeline includes calculation of perfusion parameters, automated detection of the infarct-core and the penumbra, calculation of an ASPECT-score and presentation of the results in a comprehensive report sheet showing all relevant parameters at one glance. Methods and Implementation: The method is fully implemented in MATLAB and consists of the following steps, similar to [1]. 1. Calculation of the perfusion parameters: The Artetial Input function is automatically determined based on a cluster analysis [2].

123

Features illustrated at the exhibit: All features of the software will be presented, including details on the processing pipeline and discussion of the generated report sheet (see Figure 1). Interested visitors are encouraged to bring their own data (DSC Perfusion timeseries and ADC maps in DICOM or NIFTI Format) in order to demonstrate a live demo. References: 1. Straka M, Albers GW, Bammer R. Real-time diffusion-perfusion mismatch analysis in acute stroke. Journal of Magnetic Resonance Imaging. 2010;32(5):1024–1037. 2. Kim Mouridsen, Soren Christensen, Louise Gyldensted, and Leif Ostergaard. Automatic selection of arterial input function using cluster analysis. Magn Reson Med, 55(3):524–531, 2006. 3. Automated Infarct Core Volumetry Within the Hypoperfused Tissue: Technical Implementation and Evaluation. Kellner E, Reisert M, Kiselev VG, Maurer CJ, Beume LA, Urbach H, Egger K. J Comput Assist Tomogr. 2016 Dec 15. [Epub ahead of print]. 4. Barber PA, Demchuk AM, Zhang J, Buchan AM. Validity and reliability of a quantitative computed tomography score in predicting outcome of hyperacute stroke before thrombolytic therapy. ASPECTS Study Group. Alberta Stroke Programme Early CT Score. Lancet 2000; 355: 1670–1674.

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

S691

738 An online service for fully-automatic prostate MRI segmentation G.A. Garcı´a Ferrando1, J. Juan-Albarracı´n1, E. Fuster-Garcia1, C. Monserrat Aranda2, J.M. Garcı´a-Go´mez1 1 Instituto de Aplicaciones de las Tecnologı´as de la Informacio´n y de las Comunicaciones Avanzadas (ITACA), Universitat Polite`cnica de Vale`ncia, Vale`ncia/SPAIN, 2Departamento de Sistemas Informa´ticos y Computacio´n (DSIC), Universitat Polite`cnica de Vale`ncia, Vale`ncia/SPAIN Purpose of the software: Manual prostate volume delineation is a time consuming task required for patient management, such as for radiotherapy planning and track disease progression. Moreover, this step is previously required by other image analysis tasks such as to initialize multimodal registration algorithms (1) and to obtain the region of interest for prostate cancer detection. In this work we present a prostate segmentation software integrated in the MTSImaging platform. The purpose of this work is to provide an online free accesible service for prostate segmentation competitive with the State-of-the-Art (2). The system works with T2-weighted MR images, without needing additional information. A full exam segmentation takes approximately 30 s, and the systems returns the segmentation and a posterior probability map. Methods and Implementation: The full-automatic prostate segmentation service based on T2 MRI is available in http://www.mtsimaging.com/prostate. It is implemented as a ResConvolutional Neural Network composed by 16 Encoder-Decoder layers and a final convolutional layer with a softmax function to obtain the posterior probability maps of each class as output. This schema is inspired in the U-Net architecture (3). The Neural Network has been trained using 50 MRI images from the PROMISE12 MICCAI Prostate Challenge corpus (4). The training steps were (1) image resizing to [width, height, depth], (2) we apply Contrast Limited Adaptive Histogram Equalization (5) for high intensity artifact correction, (3) robust zScore normalization using the median and the IQR. Finally, the model was trained until convergence using Adam optimizer and continuous DICE as a loss function (6), which is robust to binary unbalanced classes. We evaluate our model with a test set of 5 samples extracted from the supervised images. In addition, we evaluate the model in two different scenarios, one where the model input is the whole volume (FullAutomatic), and another where we feed the model with a manual cropped image of the prostate (Semi-Automatic). Table1. Performance of the model in both scenarios Method Dice Coefficient [%] Full-Automatic Res-CNN 82,97 Semi-Automatic Res-CNN 86,63

Features illustrated at the exhibit: The Prostate segmentation service can work directly by uploading T2 images in Dicom or NifTi formats. From that, the system automatically applies the Res-CNN network to calculate the segmentation, that is given to the user as image mask in NifTi format (see Figure 1 for details). Furthermore, the user is reported with the posterior probabilities maps of the segmented volume and information of the network and training set.

References: [1] Y. Hu, H. U. Ahmed, Z. Taylor, C. Allen, M. Emberton, D. Hawkes, and D. Barratt, ‘‘MR to ultrasound registration for imageguided prostate interventions,’’ Medical Image Analysis, vol. 16, no. 3, p. 687–703, 2012. [2] L. Yu, X. Yang, H. Chen, J. Qin, and P.-A. Heng, ‘‘Volumetric convnets with mixed residual connections for automated prostate segmentation from 3d mr images,’’ PROMISE12 MICCAI Prostate Challenge, 2016. [3] O. Ronneberger, P. Fischer, and T. Brox, ‘‘U-net: Convolutional networks for biomedical image segmentation,’’ arXiv, 2015. [4] ‘‘MICCAI Grand Challenge: Prostate MR Image Segmentation,’’ The Medical Image Computing and Computer Assisted Intervention Society, 2012. [5] K. Zuiderveld, ‘‘Contrast limited adaptive histogram equalization,’’ Academic Press Professional, vol. Graphic Gems IV, pp. 474–485, 1994. [6] F. Milletari, N. Navab, and S.-A. Ahmadi, ‘‘V-Net: Fully convolutional neural networks for volumetric medical image segmentation,’’ arXiv, 2016.

739 MTSimaging: multiparametric image analysis services for vascular characterization of glioblastoma E. Fuster-Garcia, J. Juan-Albarracı´n, J.M. Garcı´a-Go´mez Instituto de Aplicaciones de las Tecnologı´as de la Informacio´n y de las Comunicaciones Avanzadas (ITACA), Universitat Polite`cnica de Vale`ncia, Vale`ncia/SPAIN Purpose of the software: MTSimaging is an online platform that provides automated multiparametric image analysis services for the characterization of solid tumours [1]. The MTSimaging currently provides two main services focused on the vascular characterization of preoperative glioblastomas (GBM): (1) the anatomical GBM analysis, where the anatomical MR images (T1, T1c, T2, and FLAIR) are combined to generate an automated segmentation of the tumour tissues i.e. oedema, enhancing tumour, and necrosis; and (2) the Hemodynamic Tissue Signature (HTS), where the system provides a new nosologic map of the GBM tissues, grouped in four vascular habitats with different hemodynamic behaviour: the high angiogenic enhancing tumour (HAT), the low angiogenic enhancing tumour (LAT), the potentially tumour infiltrated peripheral oedema (IPE), and the vasogenic peripheral oedema (VPE). The MTSimaging technology is patented under the procedure P201431289. Methods and Implementation: MTSimaging technology offer 4 modules: the MRI preprocessing module, the anatomical segmentation module, the perfusion quantification module and the vascular habitats detection. The anatomical pipeline comprises the first two modules while the HTS service encloses all modules generating a

123

S692 richer functional segmentation (see Figure 1). The pre-processing module consist of the following steps: (1) Resampling at 1 9 191 mm3, (2) Denoising, (3) Bias field correction, (4) Registration and (5) Skull-stripping. The perfusion module quantifies the Cerebral Blood Volume (CBV), Cerebral Blood Flow (CBF) and Mean Transit Time (MTT) biomarker maps from the DSC sequence. To do so, the arterial input function is automatically computed through a divide-and-conquer approach, by selecting those curves with highest area under the curve, earliest Time To Peak (TTP), highest peak height and quickly wash-out. Finally, both anatomical segmentation and vascular habitats detection modules are based on the methods proposed in [2, 3]. These modules employ the Directional Class Adaptive Spatially Varying Finite Mixture Model (DCA-SVFMM) algorithm [4], and a hierarchical strategy to identify the pathological structures from most general to most specific.

Features illustrated at the exhibit: The MTS imaging services are freely available for non-commercial research purposes at http://www.mtsimaging.com. Both services provide a report in pdf format (see Figure 2), the pre-processed original images, the brain extraction intra-cranial mask and the GBM tissue segmentation mask. As an addition, the HTS service also provides the perfusion maps, and the HTS segmentation masks. The analysis of a study takes approximately 50 min.

123

ESMRMB Congress (2017) 30 (Suppl 1): S501–S692

References: [1] Juan-Albarracı´n, J., Fuster-Garcia, E., Garcı´a-Go´mez, J.M., An Online Platform for the Automatic Reporting of Multi-parametric Tissue Signatures: A Case Study in Glioblastoma, in: Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Presented at the International Workshop on Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, Springer, Cham, pp. 43–51. (2016). [2] Juan-Albarracı´n, J., Fuster-Garcia, E., Manjo´n, J.V., Robles, M., Aparici, F., Martı´-Bonmatı´, L., Garcı´a-Go´mez, J.M.: Automated glioblastoma segmentation based on a multiparametric structured unsupervised classification. PLOS One 10(5), 1–20 (2015). [3] Juan-Albarracı´n, J., Fuster-Garcia, E., Garcı´a-Go´mez, J.M.: Glioblastoma tissue guided segmentation through unsupervised structured classification. In: Proceedings of II International Symposium on Clinical and Basic Investigation in Glioblastoma, vol. 1, no. 3, p. 101 (2015). [4] Nikou, C., Galatsanos, N.P., Likas, C.L.: A class-adaptive spatially variant mixture model for image segmentation. IEEE Trans. Image Process. 16(4), 1121–1130 (2007).