Int J CARS (2014) 9 (Suppl 1):S7–S13 DOI 10.1007/s11548-014-1013-0
CT AND PET-CT
Correlation of FDG PET–CT quantitative parameters with serum tumor markers in recurrent colorectal cancer M. Caglar1, M. Unlu2, C. Yener1, E. Karabulut3 1 Hacettepe University Medical Faculty, Nuclear Medicine, Ankara, Turkey 2 Gazi University, Nuclear Medicine, Ankara, Turkey 3 Hacettepe University, Biostatistics, Ankara, Turkey Keywords Colon cancer FDG PET–CT Total lesion glycolysis Purpose Positron emission tomography–computed tomography (PET–CT) with 2-deoxy-2-[18F]fluoro-D-glucose (18F-FDG) is an established imaging modality which is used in the management and follow-up of colorectal cancer (CRC). Among the PET parameters, maximum standardized uptake value (SUVmax) which reflects tumor glucose metabolism is the most commonly used semi-quantitative index of 18 F-FDG metabolic rate [1]. However, the measurement of SUVmax has been confined to detection of the most obvious metabolic activities of the tumor at a single site, but not the overall tumor activity. SUVmean is the average value generated from the entire tumor, but differences in operator contouring of tumor will yield varying values. In addition, both SUVmax and SUVmean represent only the metabolic activity per gram of tissue, but they are not able to reflect tumor dimensions and volume. In contrast, total lesion glycolysis (TLG) is a newly proposed tumor marker in which both tumor activity and volume are integrated [2]. TLG is the product of SUVmean and metabolic tumor volume (MTV) which is derived from PET imaging studies. Patients with CRC can have single or multiple lesions depending on the stage of the disease. In order to take the number of lesion into consideration, SUVsum (summation of SUVmax for all tumors), whole-body MTV (MTVwb; summation of MTV for all tumors) and whole-body TLG (TLGwb; summation of TLG for all tumors) were calculated in the present study and correlated with tumor markers. The purpose of this study is (1) to investigate the sensitivity and specificity of serum tumor markers and FDG PET–CT, (2) determine the correlation of these markers with FDG PET–CT quantitative indices (SUVmax, MTV, TLG) in patients suspected to have recurrent colorectal cancer (CRC). Methods FDG PET–CT imaging was performed in 212 patients with possible CRC recurrence. Eligibility requirements for this study were as follows: (1) patients with a history of CRC who went into complete remission after treatment, (2) pathology of adenocarcinoma, (3) elevation of tumor marker (cancer antigen 19-9 and/or carcinoembryonic antigen) or suspicious radiological evaluation during follow-up after complete remission. All patients underwent an integrated PET/CT (GE Medical Systems Discovery ST 4 slice PET–CT scanner, LLC 3000 N, Grandview Blvd., Waukesha WI, USA) examination. Interpretation and semiquantitative was carried out on Advantage 4.6 workstation. (GE AW4.6 Milwaukke, Wisconsin USA). PET/CT images were interpreted as no metabolic evidence of recurrence when there was no abnormal FDG uptake in any region.
Any focus of FDG uptake with SUVmax higher than that of the surrounding tissue not related to physiologic uptake was considered as abnormal for visual analysis. For other semi-quantitative indices (MTV and TLG) any focus of FDG uptake outside normal physiologic uptake with SUVmax greater than 1.5 liver SUVmean plus 2.5 standard deviation of liver SUV was considered a measurable lesion. We used relative threshold method for measuring these semiquantitative indices and manually drew a 3D contour around voxels that were equal or greater than 42 % of the maximum voxel [3]. From each PET–CT image, SUVmax, SUVmean and MTV were extracted. TLG was calculated as the sum of the product of SUVmean and MTV of all lesions (TLG = sum of [SUVmean x MTV)]. Figure 1 gives a representative example of one of the scans in the study. Serum levels of tumor markers were obtained within 3 months of PET–CT. All enrolled cases showed elevation of a tumor marker over the reference value on at least two serial measurements before PET– CT. The statistical analyses were performed using SPSS software. Data were analyzed retrospectively. T test was used to compare functional imaging and serum tumor markers. The Spearman’s correlation coefficient was used to evaluate the correlations between the tumor markers and SUVmax, MTVwb, or TLGwb. A statistically significant difference was defined as a P value of less than 0.05. Results Two hundred and twelve patients underwent FDG PET–CT for the assessment of suspected recurrent CRC between July 2011 and July 2013. Following exclusions 57 patients with missing data or lost to
Fig. 1 F18 FDG PET–CT shows hypermetabolic lesion in the liver. PET tumor volume was calculated by drawing ROIs over axial slices containing FDG PET uptake that were equal or greater than 42 % of the maximum voxel
123
S8 follow-up, 155 patients (87 men, mean age 61 years) remained for final analysis. 97 (63 %) were referred to PET–CT imaging for elevated serum tumor markers and 53 (37 %) were referred for abnormal radiological evaluation. One hundred and fourteen (74 %) of patients had abnormal foci FDG uptake on PET–CT (60 % in the liver, 15 % involving the anastomosis site, 15 % in lymph nodes and 12 % involving distant organs). At an adaptive threshold of 42 %, the average SUVmax, SUVmean MTV and TLG of these lesion were 10.2, 5.9, 57.4 cm3 and 55.3 respectively. All FDG PET–CT most quantitative parameters correlated positively with serum CEA levels and the correlation coefficients were 0.56, 0.55 and 0.70 for SUVmax, MTV and TLG (p \ 0.0001) Serum CEA and CA 19-9 had a sensitivity of 75 and 35 % and specificity of 81 and 84 % for the detection recurrent CRC respectively. The overall sensitivity and specificity of FDG-PET for the detection of recurrent CRC in patients with high prior probability was 96 and 100 %. Sensitivity was low for patients with mucinous adenocarcinoma. Conclusion Although based on EORTC criteria, the standardized 18F-FDG uptake value of the hottest voxel in the lesion (SUVmax) has traditionally been chosen as a hallmark of tumor metabolism, choosing functional lesion parameters such as metabolic volume or total lesion glycolysis may overcome the evident limitations of SUVmax. PET–CT is more accurate than serum CA-19-9 and CEA assay in detecting recurrent CRC. Majority of the recurrences were in the liver and the sensitivity is affected by tumor histology. TLG is the most reliable FDG PET–CT index to estimate tumor activity. References [1] Mehta G, Chander A, Huang C, Kelly M. Fielding P (2014) Feasibility study of FDG PET/CT-derived primary tumour glycolysis as a prognostic indicator of survival in patients with nonsmall-cell lung cancer. Clinical Radiology 69(3) 268–274 [2] Van de Wiele C, Kruse V, Smeets P, Sathekge M, Maes A (2013) Predictive and prognostic value of metabolic tumour volume and total lesion glycolysis in solid tumours. Eur J Nucl Med Mol Imaging 40 (2): 290–301 [3] Wahl RL, Jacene H, Kasamon Y, Lodge MA (2009) From RECIST to PERCIST: Evolving Considerations for PET response criteria in solid tumors. J Nucl Med. 50 (Suppl 1):122S–150S.
Int J CARS (2014) 9 (Suppl 1):S7–S13 example, PET imaging can highlight the metabolically active regions of a tumor, which can then be targeted during biopsy. By fusing PET images with CT images during biopsy, more precise tumor screening could be enabled. However, using fused PET/CT images for imageguided biopsies also suffer from problems related to the physical acquisition of PET images. Patient respiratory motion leads to blurring and/or artifacts in PET images which, in turn, can lead to incorrect tumor localization and, thus, potentially incorrect cancer staging. In this abstract, we describe an approach for image-guided biopsy using lowradiation dose amplitude gated phase-matched PET/CT images. This approach leads to a lower radiation dose for the patient while obtaining sharper PET/CT images for biopsy guidance. Methods Our proposed clinical workflow for PET/CT guided biopsy is shown in Fig. 1. In this workflow, we first obtain the pre-operative images in the PET–CT suite. These images are obtained by using a low-radiation approach. In this approach, first the gated PET images are obtained. The amplitude level of the PET gating is then used to obtain phase matched amplitude gated prospective CT images. During this CT acquisition, the patient is exposed to radiation only during the amplitude gating. Thus, the total radiation exposure for the patient is reduced. The main advantage of acquiring gated amplitude matched PET/CT images is that motion artifacts are significantly reduced. The proposed workflow is identical to that of a standard PET/CT. In our approach, we use a peak-inspiration phase for all the gated PET/CT images. During the biopsy procedure, the intra-operative CT images are also obtained at peak inspiration. Thus, the pre-interventional PET images can be immediately fused with the CT images to provide tumor localization during biopsy. Results In this section, we present the results of our approach by using an Anzai respiratory gating phantom. This phantom consists of a piston that moves at a constant speed. In our test, we set this speed of motion to 12 cycles per minute. A line source was attached on top of the cylinder so that it is visible in both CT and PET images. We used a Siemens mCT 64 detector PET/CT machine with the VG40 version of the Siemens software. During the image acquisition, we collected four sets of data: (1) attenuation-correction CT; (2) regular PET; (3) Gated PET; and (4) gated prospective CT. In Fig. 2, we see that the regular PET/CT fusion suffers due to the cylinder motion. Thus, the CT image of the line detector is curved. Also the pixels depicting the functionally active part are more dispersed in the PET image as
Evaluation of a method using low-radiation dose amplitude gated phase-matched PET–CT images for guided liver biopsy R. Khare1, K. Cleary1, A. Enquobahrie2, P. Kinahan3, G. Esposito4, F. Banovac4 1 Children’s National Medical Center, Washington, DC, USA 2 Kitware Inc., Carrboro, USA 3 University of Washington, Seattle, USA 4 Georgetown University, Washington, DC, USA Keywords PET/CT Image-guided surgery Liver biopsy Percutaneous biopsy Purpose Percutaneous biopsy is a common procedure for staging of cancer. In this procedure, a biopsy needle is inserted through the skin to obtain samples of the suspicious nodule to test for malignancy. This procedure is often carried out under computed tomography (CT) guidance. Sometimes malignant lesions are not well detected or characterized by CT, and the incorporation of a molecular imaging modality such as positron emission tomography (PET) can enable biopsy of functionally active tissue, instead of relying solely on anatomical information. For
123
Fig. 1 Clinical workflow for motion-free PET/CT guided biopsy
Int J CARS (2014) 9 (Suppl 1):S7–S13
S9 The determinant of the Jacobian measures the change in volume that occurs due to the deformation of the lung. The registration between expiratory and inspiratory scans is performed using a diffeomorphic elastic registration implemented in ANTs [4]. The registration has three steps: linear translation, affine registration and diffeomorphic elastic registration following the approached proposed in [5]. The residual image is then defined as
Fig. 2 Fused PET/CT images. Left fused attenuation corrected CT and regular PET images. Right fused amplitude gated phase matched PET/CT images
compared to the sharp amplitude-gated phase matched PET/CT image where the functionally active image pixels closely surround the outline of the line detector. Conclusion We have presented an approach for low radiation motion-free PET/ CT guided biopsy. Our preliminary results with a moving Anzai phantom show the qualitative improvement of our approach over the standard PET/CT image protocols. After further testing, we plan to begin a clinical trial using this technique. IRB approval for this clinical trial has already been obtained. Acknowledgements This work was supported by NIH/NCI grant R42EB000374.
A quantitative method for mosaic gas trapping based on residual mass R. San Jose´ Este´par1,2, A. Reeves3, D. Yankelevitz4, C. Henschke4, D. Mendelson4, R. de La Hoz4 1 Brigham and Women’s Hospital, Radiology, Boston, USA 2 Harvard Medical School, Boston, USA 3 Cornell University School of Engineering, Ithaca, USA 4 Icahn School of Medicine at Mount Sinai, New York, USA Keywords Air trapping Biomarker Computed tomography Registration Purpose Air trapping is a pathophysiologic condition indicating the retention of excess gas in all or part of the lung at any stage of expiration. Expiratory CT has been used to reveal air trapping in patients with airway diseases such as pulmonary emphysema, chronic bronchitis, asthma, and small-airway disease. Air trapping can be classified as extensive and mosaic. Extensive gas trapping is characterized by well-defined low attenuation areas in expiratory scans. Mosaic air trapping is characterized by the appearance of inhomogeneous air emptying on expiratory compared to normal inspiratory chest CT scans. A number of quantitative chest CT methods have been proposed to assess extensive gas trapping [1, 2], however those methods fail to detect the diffuse nature of mosaic gas trapping. We have developed an approach to tackle this problem and we have validated the technique in a population of World Trade Center workers with abnormal lung emptying reported by a radiologist [3]. Methods The new quantitative approach is based on measuring the residual mass between the inspiratory scan and the expiratory scan registered into the inspiratory scan reference frame. When heterogeneous emptying of the lung occurs, the residual mass between the expiratory and inspiratory scans that cannot be accounted by the change in volume that take place can be used a metric for the amount of mosaic air trapping that may have occurred. The residual mass image is computed by subtracting the densities of the registered expiratory scan from the inspiratory scan and multiplying by the determinant of the Jacobian of the deformation between expiration and inspiration.
Residual Mass ¼ ðIinsp ðxÞ ðT o Iexp ðxÞÞjJT ÞjJT j where Iinsp and Iexp are the inspiratory and expiratory scans respectively, T is the concatenated transformation and JT is the Jacobian of the transformation. The residual mass percentage (RMperc) is defined as the percentage of the lung whose residual mass is above a given threshold. For this study, we define RMperc100 as the percentage of the lung whose residual mass is above 100 mg. Our approach was validated using a cohort of 24 former WTC workers. A radiologist visually assessed these subjects and the extent of abnormal air trapping was reported using a 0–4 scale for each lung region. A total scored was computed by summing the regional scores. We reviewed a previously reported set of paired view chest CT scans on 29 former WTC workers [3]. Our method was compared against three other methods to assess extensive air trapping: (1) percentage of the lung whose expiratory attenuation was less than 856 HU (AT856) [2], and (2) the recently proposed functional small airway disease (fSAD) [1]. Our comparison was done at a global level (All) and for the upper, middle and lower lung regions. Results Figure 1 shows the residual mass map computed by the proposed method corresponding to one case with mosaic gas trapping. The correlation between the expert visual score for air trapping and each metric for each lung region is shown in Table 1. RMperc100 achieves the best positive correlation. It is worth noting that both AT856 and fSAD have a negative correlation with the visual score suggesting that these previously proposed quantities may not be suitable for mosaic gas trapping. Figure 2 shows the mean metric value for each visual score level. RMpec100 mean value increases with increase level of severity. However, AT856 and fSAD show an inverse trend. All the methods report a mean value greater than 10 % for a baseline visual score of 0 suggesting that the methods are more sensitive to sublet changes than the visual assessment performed by the radiologist.
Fig. 1 Residual map image corresponding to a case with mosaic air trapping. More blue indicates a greater residual mass disparity suggesting air trapping
Table 1 Correlation between the expert visual score for air trapping and each metric for each lung region Metric RMperc100
All 0.744
Upper 0.723
Middle 0.851
Lower 0.579
AT856
-0.587
-0.44
-0.641
-0.49
fSAD
-0.579
-0.412
-0.638
-0.562
123
S10
Fig. 2 Mean air trapping metric for each radiologist visual score level Conclusion We have introduced a method to objectively assess mosaic gas trapping based on the mass discrepancy between inspiratory and expiratory CT scans that cannot be explained just by volume change. The results of approach show that our method is highly correlated (greater than 0.7) with visual scoring done by a radiologist. In contrast, current proposed methods to quantify extensive air trapping show a negative correlation suggesting that they may not be adequate to capture mosaic air trapping. More work is needed to elucidate the pathophysiological meaning of the residual mass approach. Additional work is needed to quantify the range of normality for residual mass. References [1] Galba´n CJ, Han MK, Boes JL, Chughtai KA, Meyer CR, Johnson TD, Galba´n S, Rehemtulla A, Kazerooni EA, Martinez FJ, Ross BD. Computed tomography-based biomarker provides unique signature for diagnosis of COPD phenotypes and disease progression. Nat Med 2012;18:1711–1715. [2] Mets OM, Zanen P, Lammers J-WJ, Isgum I, Gietema HA, van Ginneken B, Prokop M, de Jong PA. Early identification of small airways disease on lung cancer screening CT: comparison of current air trapping measures. Lung 2012;190:629–633. [3] Mendelson DS, Roggeveen M, Levin SM, Herbert R, la Hoz de RE. Air trapping detected on end-expiratory high-resolution computed tomography in symptomatic World Trade Center rescue and recovery workers. J Occup Environ Med 2007;49:840–845. [4] Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage 2011;54:12–12. [5] Murphy K, van Ginneken B, Reinhardt JM, Kabus S, Ding K, Deng X, Cao K, Du K, Christensen GE, Garcia V, Vercauteren T, Ayache N, Commowick O, Malandain G, Glocker B, Paragios N, Navab N, Gorbunova V, Sporring J, de Bruijne M, Han X, Heinrich MP, Schnabel JA, Jenkinson M, Lorenz C, Modat M, McClelland JR, Ourselin S, Muenzing SEA, et al. Evaluation of registration methods on thoracic CT: the EMPIRE10 challenge. IEEE Trans Medical Imaging 2011;30:1901–1920.
Int J CARS (2014) 9 (Suppl 1):S7–S13 Purpose Optical computed tomography (CT) obtains a tomographic image using visible ray and near infrared ray unlike X-ray CT. Therefore, there are many advantages against X-ray CT. For example, there is no risk of radiation exposure. Furthermore, it is possible to design compact and inexpensive equipment. Optical-CT can be classified into two categories: one uses the reflection information and one uses transmission information [1]. In this study, we focused on the opticalCT employing the transmission information using visible ray. It provides a tomographic image using reconstruction algorithm similar to that of X-ray CT. This technique is intended for mainly inorganic such as the gel dosimeters used in radiotherapy [2], and objects that visible light is easily to transmit such as cells and tumors [3]. In particular, optical-CT for micro-object such as cells and tumors is called micro optical-CT. Since it is possible to obtain three-dimensional image easily, as well as undistorted, compared with tomographic image by the conventional sectioning method. However, existing micro optical-CTs output monochrome images; useful color information that we use at the time of observation has not been obtained. Therefore, the main purpose of this study was to develop a color micro optical-CT system to obtain the colored tomographic images. Furthermore, we also evaluated proposed system using color phantom and biological samples. Methods Our experimental system (Fig. 1) consists of digital single-lens reflex camera (Canon EOS 60D) and a trinocular microscope (Carton, DSZT-44FT). Observable sample size is in the range of 1.7–7.0 mm at the viewpoint of the depth of field. Magnification of the microscope can be changed between 910 and 944, and total pixel resolution of camera connected to the microscope can be changed between 0.7 and 3.1 lm by changing magnification. The sample was fixed in a test tube by the cellulose solution. The test tube was connected to a step motor; it was rotated automatically by motor driver. In order to obtain color slice image, the projection images were obtained from various projection angles by rotating the test tube. And then, the projection images were decomposed R, G, B components, and performed image reconstruction for each component using filtered back projection method. Finally, color tomogram is obtained by combining threecolor component images. Here, these processing, were automated by computer program that was developed by Microsoft Visual C++ 2010. Results In the experiment, color phantom and color biological sample excellent in light transmission characteristics were scanned by our experimental system. Using phantom, the reproducibility of color and
Preliminary study on color micro optical-CT: evaluation of experimental system using biological samples C. Murata1, A. Teramoto1, H. Fujita2 1 Fujita Health University, Graduate School of Health Sciences, Toyoake-city, Aichi, Japan 2 Gifu University, Department of Intelligent Image Information, Division of Regeneration and Advanced Medical Sciences, Graduate School of Medicine, Gifu-City, Gifu, Japan Keywords Biomedical Optical CT Microscope Color
123
Fig. 1 Color micro optical-CT
Int J CARS (2014) 9 (Suppl 1):S7–S13 shape of the tomogram were evaluated. It was found that the color and shape of the tomogram obtained by the color micro optical-CT system were similar to samples. As for the evaluation using biological sample, we obtained the clear color tomographic image of it. Conclusion In this study, we aimed to develop the color micro optical-CT as a novel imaging technique. In the experiments, we evaluated the effectiveness of this method by use of color phantom and biological samples. As a result, internal shape and color information of samples were obtained. Furthermore, the reproducibility of them was also good. These results indicate that our color micro optical-CT may be useful for the color analysis of small samples. References [1] Fujita H, Ishida T, Katsuragawa S, ‘‘Handbook of Practical Image Analysis in Medicine,’’ Ohomsha, Tokyo, 2012, in Japanese. [2] Gorge JC, Ranade M, Maryanski MJ, Schulz RJ, ‘‘Radiation dose distribution in three dimensions from tomographic optical density scanning of polymer gels: I. Developed of an optical scanner,’’ Physics in Medicine and Biology, vol. 41, no. 12, pp. 2695–2704, 1996. [3] Oldham M, Sakhalkar H, Olover T, Wang YM, Kirkpatrick J, Cao Y, Badea C, Johnson GA, Dewhirst M, ‘‘Three-dimensional imaging of xenograft tumors using optical computed and emission tomography,’’ Medical Physics, vol. 33, vol. 9, pp. 3193–3202, 2006.
Scintigraphic bremsstrahlung and PET-derived assessment of Yttrium-90 lung shunt fraction using a multi-compartment phantom model C. Wright1, R. Layman1, M. Tweedle1, M. Knopp1, N. Hall1 1 Ohio State University Wexner Medical Center, Radiology, CO, USA Keywords PET/CT Quantification Theranostics Interventional oncology Purpose Yttrium-90 (90Y) microsphere radioembolization remains an option for patients with unrespectable liver malignancies and metastases. Most of the radioactivity produced by 90Y results from the production of therapeutic electrons and locally kill the targeted liver lesions. 90Y also results in the production of bremsstrahlung radiation that can be detected using standard gamma cameras. Scintigraphic bremsstrahlung imaging has been qualitatively used by many institutions to verify the microsphere distribution within the targeted hepatic lesions as well as assess for any significant extrahepatic deposition which could result in damage to normal non-targeted tissues. It has been recently demonstrated that PET/CT imaging of the annihilation radiation generated by 90Y internal pair production can produce better resolution images than scintigraphic bremsstrahlung imaging and offers quantitative capabilities. There are currently no available studies that evaluate or compare these imaging techniques for assessment of post-therapy lung shunt fraction. This study evaluated both scintigraphic bremsstrahlung and PET/CT imaging using a multi-compartment phantom model of lung shunt fraction simulating a post-therapy 90Y radioembolization patient. Methods This model consists of two large phantoms placed adjacent to each other. One phantom simulates the liver and contains a hollow sphere with higher 90Y activity to simulate a liver lesion. The other phantom simulates the lungs. 90Y salt solution (925–1,110 MBq of total 90Y activity obtained from the University of Missouri Research Reactor) was placed into each of these 3 compartments in different proportions to simulate patients with different lung shunt fractions and varying tumor to background ratios. Serial imaging studies were
S11 performed using conventional PET/CT and scintigraphic bremsstrahlung imaging. Using the Fluorine-18 (18F) imaging protocols, 10, 15 and 30 min PET acquisition were performed. No effort was made to prevent saturation of the PET detectors from bremsstrahlung radiation. Results Preliminary results verify that multi-compartment 90Y phantoms can be imaged using scintigraphic bremsstrahlung (i.e., portable gamma camera, stationary gamma camera, SPECT, and SPECT/CT) and PET/CT approaches and provide diagnostic images. Quantitative measurements can be obtained to estimate the simulated lung shunt fraction. PET-derived mean 90Y activities and volumes for the different compartments were calculated over a time period corresponding to 3–4 90Y half-lives and compared to those derived from bremsstrahlung SPECT/CT. System detector sensitivity as well as qualitative and quantitative multi-compartment phantom assessment obtained with varying acquisition times and different imaging modalities were also evaluated. Conclusion Our preliminary results verify that 90Y activity can be imaged and quantified with conventional imaging systems. This imaging approach can be used to provide more accurate data on the distribution of 90Y microspheres following radioembolization for the purposes of patient safety and quality control. 90Y radioembolization is anticipated to increase in the future and likewise rare adverse complications like radiation pneumonitis may become more frequent. Recognition of such potential complications, as well as striving to improve the diagnostic and prognostic stratification of radioembolization patients before and after therapy, is paramount for interventional and diagnostic radiologists, radiation oncologists, nuclear medicine physicians and medical/surgical oncologists. Continued development of new computer-assisted methodologies in quantitative PET/CT and SPECT/CT imaging will have broader clinical applications to a wide variety of cancer types which may be amendable to radio-guided surgical intervention or novel peptide-, antibody- or microspherebased radiotherapies.
Development of marker-free estimation method of colonoscope tip position using electromagnetic sensors and CT volumes H. Kondo1, M. Oda1, K. Furukawa2, R. Miyahara2, Y. Hirooka3, H. Goto2, T. Kitasaka4, K. Mori5 1 Nagoya University, Graduate School of Information Science, Nagoya, Japan 2 Nagoya University, Graduate School of Medicine, Nagoya, Japan 3 Nagoya University Hospital, Department of Endoscopy, Nagoya, Japan 4 Aichi Institute of Technology, School of Information Science, Toyota, Japan 5 Nagoya University, Strategy Office, Information and Communications Headquarters, Nagoya, Japan Keywords CT images Colonoscope tracking Electromagnetic sensor Line registration Purpose In recent years, CT colonography (CTC) become popular as new diagnostic method of the colon in the clinical field. When a physician finds polyps during CTC, a patient will be referred to endoscopy. Polypectomy will be performed to remove polyps found in CTC under an endoscopic procedure. However, it is difficult to find the polyps during colonoscopy. This is because the colon has a long and winding shape with many haustra. Endoscopic views are quite similar. Development of colonoscope navigation system is expected to be developed. This system localizes the current location of the colonoscope tip and guides a physician to the location of polyps found in CTC.
123
S12 Previous method of tracking a colonoscope registered a sensor coordinate system to a CT coordinate system by a point-based rigid registration method [1]. However, this method needs manual interaction to obtain correspondence between markers on a patient body and those on a CT volume. Therefore, we develop a method for registering the two coordinate systems without any markers. Methods Overview In the proposed method, the colonoscope shape is traced by multiple positional trackers which insert into the colonoscope’s working channel. We calculate a curve representing the shape of a colonoscope (colonoscope line) utilizing electromagnetic (EM) sensors and a curve that represents the shape of the colon (colon centerline) extracted from a CT volume. Coordinate systems are roughly registered by the ICP algorithm. After registering a sensor and CT coordinate systems, the colonoscope tip position is estimated based on a line registration between the colonoscope line and colon centerline. Extraction of the colon centerline from CT volume Colon region is extracted from a CT volume based on the thresholding. We apply the thinning and the smoothing processes to the colon region to obtain a colon centerline (Fig. 1c). The colon centerline is represented as a set of points on the curve p[i] (i = 1, …, I). Estimation of the colonoscope line using EM sensors We insert an Aurora 5/6DOF Shape Tool, which has seven EM trackers, into the colonoscope’s working channel. We estimate a colonoscope line that represents the colonoscope shape from the sensor outputs. The Hermite spline interpolation is utilized for colonoscope line estimation. The colonoscope line is represented as a set of points on the curve q[j] (j = 1, …, J). Coordinate system registration We calculate the transformation matrix that represents the relation between the sensor and CT coordinate systems. The ICP algorithm is employed here. We assume that the proposed method is applied when a colonoscope inserted to the cecum (Fig. 1b). This means that the proposed navigation system starts when the colonoscope is fully inserted into the colon. We register between points on the colon centerline and points on the colonoscope line. However, it is difficult to register the colonoscope line and the colon centerline by the ICP algorithm because the colon centerline deforms largely in colonoscopic polypectomy. To solve this issue, we generate a deformed colon centerline obtained by deforming the colon centerline locating at deformable sections of the colon (Fig. 1d). We register the deformed colon centerline to the colonoscope line and obtained the transformation matrix between the CT and the sensor coordinate systems.
Fig. 1 a Colon phantom before colonoscope insertion. b Colon phantom after insertion to the cecum. c Colon centerline extracted from a. Five gray circles indicate landmarks. Landmarks divide the colon centerline into deformable (sections painted by red) and less deformable (sections painted by blue) sections. d Deformed colon centerline calculated from c. Deformed part is made by straighten the colon centerline (red line)
123
Int J CARS (2014) 9 (Suppl 1):S7–S13
Fig. 2 a Experimental setup of colonoscope tracking. b Positions of markers (green circle) on colon phantom
Line registration We find correspondence between points on the colon centerline and the colonoscope line by using landmarks and the length along the colon centerline [1]. The transformation matrix obtained in the previous step is applied to the colonoscope line. By using the correspondences, we estimate the point on the colon centerline corresponding to the colonoscope tip position. Results We measured the errors of colonoscope tip positions and the estimated tip position on the colon centerline. Accuracy evaluation is performed on the colon phantom (Fig. 2a). The Aurora 5/6DOF Shape Tool was attached in a colonoscope’s working channel. We attached six markers (M1–M6) on the colon phantom (Fig. 2b). The phantom was scanned by a CT scanner. The actual location of the tip of the colonoscope is identifiable by observation because the colon phantom is made of thin rubbers and tip light is identifiable from outside. When the colonoscope tip reaches at a marker position, we measured the distance along the colon centerline from the estimated position of the colonoscope tip on the colon centerline to the marker position projected onto the colon centerline. The measured distance is recorded as an error. We measured the errors 10 times at each marker point. The mean errors were 30.7 mm for M1, 126.9 mm for M2, 15.7 mm for M3, 33.3 mm for M4, 127.3 mm for M5, and 71.4 mm for M6. Acceptable error is less than 50.0 mm when we consider the viewing area of the colonoscope. We can say that the proposed method has enough small errors at the segments where less deformation is observed (M1, M3, and M4). However, errors at the deformable segments of the colon (M2, M5, and M6) are over 50.0 mm. This is because the proposed method does not consider colon deformations in the line registration. Conclusion This paper proposed a colonoscope tracking method by using EM sensors and a CT volume. Global transformation matrix between the CT and the sensor coordinate systems is obtained by the marker-free rigid registration method based on ICP. Then we estimated the colonoscope tip position based on the line registration. Experimental results showed that the proposed method can obtain small position estimation errors at colon segments that less deformation is observed. References [1] Oda M, Acar B, Furukawa K, Kitasaka T, Suenaga Y, Navab N, Mori K (2013) Colonoscope tracking method based on line registration using CT images and electromagnetic sensors. Int J CARS2013 8(1): S349–S351
Int J CARS (2014) 9 (Suppl 1):S7–S13 An algorithm for noise correction of dual energy computed tomography material density images R. Simon Maia1, C. Jacob1, A. K. Hara2, A. C. Silva2, W. Pavlicek2, J. R. Mitchell2 1 University of Calgary, Department of Computer Science, Calgary, Canada 2 Mayo Clinic, Department of Radiology, Scottsdale, Canada Keywords Material density Dual energy computed tomography Noise reduction Anisotropic diffusion Purpose Dual Energy Computed Tomography (DECT) images can undergo a two material decomposition process, which results in two images containing material density information. According to Alvarez et al. in [1], material density images obtained by that process result, in images with increased pixel noise. Therefore, noise reduction of those images is necessary in order to improve image quality. Methods A noise reduction algorithm for material density images was developed and tested using 3 different types of acquisition systems (GE’s single source DECT, Siemens’ dual source DECT and by usage of a
Water Density Image. Original Image is shown on top, noise-reduced image on bottom.
Iodine Density Image. Original Image is shown on top, noise-reduced image on bottom.
Fig. 1 Noise reduction in GE’s DECT images. Corrected images have higher SNR and CNR. The corrected images also have reduced speckle and sharper edges. Each material density image is shown with the same respective windowing settings
S13 Table 1 Average SNR Gain, SNR2 Gain and CNR Gain over all images evaluated Image type
SNR gain (%)
SNR2 gain (%)
CNR gain (%)
Average improvement over all images Improvement on water density images
46.72
33.56
51.01
Improvement on iodine density images
51.95
35.83
53.08
Average improvement over all images
49.33
34.69
52.05
simulation software called ImaSim [3]). A three level wavelet approach combined with the application of an anisotropic diffusion filter [4] was used. During each level, the resulting noise maps are further processed, in a similar fashion to Kalender et al. technique in [3], until the original resolution is reached and the final noise maps obtained. Our method works in image space and, therefore, can be applied to any type of material density images obtained from any DECT vendor. A quantitative evaluation of the noise reduced images using the signal to noise ratio (SNR), contrast to noise ratio (CNR) and 2D noise power spectra (NPS) was done to quantify the improvements. Results The noise reduction algorithm was applied to a set of images resulting in images with higher SNR and CNR than the raw density images obtained by the decomposition process. The average improvement in terms of SNR gain was about 49 % while CNR gain was about 52 %. The difference between the raw and filtered ROI mean values was far from reaching statistical significance (minimum p[0.89, average p[ 0.97). Figure 1 demonstrates the improvement in image quality in noise corrected images obtained from a GE single source DECT scanner. Table 1 synthesis the average improvement over all images examined. Conclusion We have demonstrated throughout a series of quantitative analyses that our novel noise reduction algorithm improves the image quality of DECT material density images. References [1] Alvarez R, Seppi E: Comparison of Noise and Dose in Conventional and Energy Selective Computed-Tomography. Ieee Transactions on Nuclear Science 26(2), 2853–2855 (1979) [2] Kalender W, Klotz E, Kostaridou L: An algorithm for noise suppression in dual energy CT material density images. Medical Imaging, IEEE Transactions on 7(3), 218–224 (1988) [3] Landry G: ImaSim, a Simulation Software Package for the Teaching of Medical X-ray Imaging (2010) [4] Perona P, Malik J: Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on Pattern Analysis and Machine Intelligence 12(7), 629–639 (1990)
123