Journal of Applied Spectroscopy, Vol. 79, No. 1, March, 2012 (Russian Original Vol. 79, No. 1, January–February, 2012)
PROTON MAGNETIC RESONANCE SPECTROSCOPY IN DEEP HUMAN BRAIN STRUCTURES AT 7 T M. Elywa*, S. Mulla-Osman, F. Godenschweger, and O. Speck
UDC 543.422.25:612.824
The increased magnetization and frequency separation at high magnetic field strength, such as 7 T, can provide spectra of high signal-to-noise ratio and spectral resolution. However, most human brain magnetic resonance spectroscopy (MRS) studies at 7 T have employed surface coils and thus limited to superficial brain structures. In this study, volume coil excitation together with volume array reception has been utilized to access deeper brain areas. RF power limitations have been addressed by the use of VERSE-modified pulses, and spectra in parietal and pregenual anterior cingulate cortex (pgACC) have been acquired in eight subjects using STEAM with a short echo time of 20 ms. Spectra were analyzed using LC-model. Therefore, an experimental basis set of in vitro spectra was established from 20 human brain metabolite solutions. An exemplary comparison with an optimized PRESS-based single voxel MRS method at 3 T has been performed. Despite the intrinsically lower signal in STEAM, the 7 T spectra show 1.87 times higher signal-to-noise ratio than at 3 T (using PRESS) and more metabolites could be quantified reliably. The results show that the proposed method can be employed at 7 T in deep brain structures and allows the absolute and relative concentrations of human brain metabolites to be determined with low error levels. Keywords: proton magnetic resonance spectroscopy, in vivo MRS, human brain, metabolites quantification, LC-model. 1
Introduction. Proton magnetic resonance spectroscopy ( H MRS) can study brain metabolism non-invasively and provide markers for specific diseases [1, 2]. Signal-to-noise ratio (SNR), spatial and spectral resolution, as well as scan time are important quality measures of MRS. A number of localization methods have been proposed. Single volume spectroscopy (SVS) using pulse sequences that are based on spin echoes (PRESS) [3] or stimulated echoes (STEAM) [4] provides the highest spectral quality, albeit the spatial information is limited. Moreover, previous studies 1 have shown that increasing the magnetic field strength increases the H NMR spectral peak separation and SNR, thus providing higher sensitivity compared to lower field strengths [5–11]. o With standard RF-amplifiers and volume RF-coils at 7 T, short 180 refocusing pulses are difficult to achieve in most parts of the brain. Previous high-field MRS studies are therefore frequently employed surface coils with locally higher B1-field amplitude [12, 13]. However, not all brain areas can be sufficiently examined with surface coils. Since the goal of this study was to examine deeper human brain areas, the STEAM volume selection principle has o o been applied. The 90 -pulses require only half the peak power of the 180 -pulses in PRESS. Moreover, STEAM with short echo times benefits from reduced signal decay due to J-evolution of coupled spin system, and thus allows more precise quantification of metabolites, especially with short T2 relaxation time [14]. Short echo times have been used in several studies [15–18]. In this work TE was selected to be 20 ms. Previous studies that employed STEAM with ultrashort echo times were again limited by surface coils that did not allow detection in deep brain regions. TEs down to 6 ms were possible and metabolites were resolved better at 7 T than at 3 T [14]. A quadrature (transmit/receive) surface RF coil or a half-volume RF coil have been used for these studies [19, 20]. In this work, a CP-transmit and a 24-channel receive head volume coil have been used to acquire spectra from deep brain areas. ∗
To whom correspondence should be addressed.
Department of Biomedical Magnetic Resonance, Otto-von-Guericke University, 44 Leipziger Str., Magdeburg, 39120, Germany, e-mail:
[email protected]. Published in Zhurnal Prikladnoi Spektroskopii, Vol. 79, No. 1, pp. 129–135, January–February, 2012. Original article submitted June 22, 2011. 120
0021-9037/12/7901-0120 ©2012 Springer Science+Business Media, Inc.
To further reduce the required RF-peak power, variable-rate selective excitation (VERSE) has been introduced into the STEAM sequence for this study. VERSE applies a time-varying gradient waveform in combination with a modified RF waveform to distribute the RF-amplitude more evenly throughout the pulse and thus reduce RF power [21]. 1 The goal of this study is to demonstrate an in vivo H MRS method [22] using a combination of STEAM with VERSE to reduce RF peak power requirement and to determine human brain metabolite concentrations in small and deep brain regions using volume coils. 7 T spectra are compared to 3 T results. Material and Methods. All measurements were performed on a 7 T whole body MRI and whole body 3 T systems (Siemens Medical Solutions, Erlangen, Germany and Siemens Magnetom Trio, respectively). At 7 T, a 24channel volume head coil with local circularly polarized volume excitation (Nova Medical, Inc. USA) was used for acquisitions in the human brain. A circularly polarized small volume coil of 16 cm inner diameter (Doty Scientific, Inc., Columbia, USA) was used for in vitro measurements of the LC-Model basis set to improve SNR due to a higher filling factor. An 8-channel volume head coil has been used at 3 T for signal reception in human measurements, whereas the body coil was used for excitation. In this work, a 24-channel volume coil was used for in vivo measurements. A big difference between the surface and the volume coils is that the volume coils are always much larger than surface coils and can completely surround a human head. Although the surface coils have higher SNR because they receive noises only from superficial brain layers, they were usually used as receive coils only. On the other hand, the volume coils are used both as receiver and transmitter coils and can acquire the signal from deep brain regions. STEAM optimized with VERSE pulses (TR/TE/TM = 3000/20/15 ms) has been used at 7 T since RF-peak power limitations prohibited the use of PRESS with volume coils at 7 T [23]. At 3 T, PRESS (TR/TE = 2500/35 ms) has been applied without the need for further modifications due to RF-power limitations. A water suppression (WS) scheme for both field strengths used three 55 Hz Gaussian pulses. o Two important parameters used to reduce the transmitted voltage for a 90 pulse are the VERSE factor (Vf), corresponding to the gradient amplitude reduction for the center RF-pulse lobe relative to a constant slice selection gradient, and the pulse duration. To test the VERSE effect, the transmitted voltage for STEAM (Vf = 1 and pulse duration 2.5 ms) and STEAM with VERSE (Vf = 1.4 and pulse duration 2.5 ms) pulses was varied from 100 to 510 V, and the unsupo pressed water signal intensity was measured. Since all three 90 pulses have been modified, the acquired signal is pro3 o portional to sin (aU), where U is the applied voltage. The voltage for the 90 flip angle has been calculated as the peak location for STEAM and STEAM with VERSE. Twenty human brain metabolite aqueous solutions, acetate (Ace, 200 mM), alanine (Ala, 100 mM), aspartate (Asp, 40 mM), citrate (Cit, 200 mM), creatine (Cr, 50 mM), gamma-amino butric acid (GABA, 200 mM), glucose (Glc, 200 mM), glutamine (Glu, 50 mM), glutamate (Gln, 100 mM), glycine (Gly, 200 mM), glycerophosphocholine (GPC, 33 mM), gluthatione (GSH, 70 mM), myoinositol (Ins, 200 mM), lactate (Lac, 100 mM), N-acetylaspartate (NAA, 50 mM), phosphocreatine (PCr, 50 mM), phosphocholine (PCho, 27 mM), phosphorylethanolamine (PE, 50 mM), succinate (Suc, 100 mM), and taurine (Tau, 200 mM), were prepared to create an in vitro basis set for the LCModel. Spherical phantoms of 40 mm in diameter made from plastic have been used as solution containers. Two hundred millimoles of Na-formate and 1 mM of 3-(trimethylsilyl) propionic-2,2,3,3-d4 (TSP) have been added to the metabolite solutions to obtain two reference peaks at chemical shifts δ = 8.4 ppm and δ = 0 ppm, respectively. All metabolites stayed in the scanner room for a few hours before the experiment was begun in order to keep them at room temperature. Twenty in vitro spectra of the basis set have been acquired at 3000 ms repetition time (TR), 15 ms mixing time (TM), 20 ms echo time (TE), 15 × 15 × 15 mm voxel size, 3600 Hz acquisition band width, and 2048 sample points. Eight healthy subjects (age 25 to 44 years) were studied after giving informed consent according to procedures approved by the local institutional review board. All in vivo human brain spectra at 7 T were measured using TR/TE/TM=3000/20/15 ms. Moreover, a volume of interest (VOI) was selected according to anatomical dimensions of the pgACC and parietal target regions. After eddy current correction based on non-water suppressed acquisitions, the absolute and relative metabolite concentrations, spectral line width, and Cramer-Rao lower bounds (CRLB) were evaluated by the LC-Model [24, 25] 121
Fig. 1. In vivo 1H NMR spectrum acquired from parietal white matter at 7 T. STEAM with VERSE (Vf = 1.4), TR/TE/TM = 3000/20/15 ms, voxel size 15 × 15 × 15 mm, BW = 3600 Hz, data points of 2048 and averages 128 (a) VOI location; (b) measured data; (c) LC-Model fit of the spectrum; (d) individual metabolite curves for each metabolite of the spectrum, and (e) fit residual. using an unsuppressed water signal from the same VOI as a reference signal. Quantifications for each metabolite are given in concentrations in mM together with the CRLB in % and full width at half maximum (FWHM) in Hz. In addition, the SNRs of all the metabolites were estimated by taking the ratios of the metabolite peak heights to twice the root mean square of the standard deviation. The standard deviation was estimated in the range from 10 to 12 ppm at 7 and at 3 T from 9 to 10 ppm (where no signals were expected). Results and Discussion. The VERSE technique with Vf = 1.4 reduced the required transmitter vol-tage by 28% compared to unmodified pulses, reducing the pulse peak power by 48%. For STEAM, this reduction allowed us o to achieve the 90 flip angle throughout the brain with the applied volume head coilà[26]. Figure 1 shows a phasecorrected spectrum measured at 7 T in the parietal brain region (Fig. 1a) by single voxel STEAM with 128 averages. A thin line is raw data and a solid line shows an LC-Model fit in Fig. 1b and c. In Fig. 1e, a small fit residual shows matching between the raw data and the fit. The narrow line width (FWHM = 0.027 ppm) of NAA indicates a good field homogeneity at 7 T. In addition, Fig. 1d shows the corresponding single metabolite fits of the spectrum. The human brain metabolites (Glu and Gln), (Cr and PCr), and (NAA and NAAG) can be separated and detected at 7 T. 122
Fig. 2. Human brain NMR spectra acquired from the pgACC region in one volunteer: (a) at 7 T using STEAM with VERSE, 128 averages, TR/TE/TM = 3000/20/15 ms, and voxel size 1.8 ml; (b) at 3 T using PRESS, 256 averages, TR/TE = 2000/35 ms, voxel size 2.5 ml, together with the LC-Model fitting curve. Figure 2 shows the human brain spectra of the same subject measured in the pgACC region at 7 T using STEAM with VERSE (TR/TE/TM = 3000/20/15 ms, Vf = 1.4, and 128 averages) and at 3 T using PRESS (TR/TE = 2000/35 ms and 256 averages) [27]. The STEAM sequence results in a 50% signal loss compared to PRESS due to the stimulated echo signal generation. Due to increased magnetization, SNR increases by a factor of 2.3 at 7 T relative to SNR at 3 T. The experimental data from three measurements at both field strengths, with fewer averages and smaller voxels, show that the SNR increases at 7 T by a factor of 1.79 compared to that measured at 3 T. When the data at both fields were corrected for the number of averages, the voxel volume, and the line width differences, the SNR increases by a factor of 1.94 when passing from 3 T to 7 T. Considering the STEAM signal loss results in a factor of 3.88. While this is higher than expected, a number of additional factors may contribute to this gain. The echo time is shorter with the STEAM sequence and the receive coil may be significantly more sensitive at 7 T due to the higher number of channels. At 7 T, the peaks of Asp in the range from 2.6 to 2.8 ppm can be clearly identified. Furthermore, the high sensitivity and increased chemical dispersion at 7 T make it possible to resolve metabolites with very low concentrations, such as Gln, GABA, and NAAG. Table 1 shows the absolute concentrations and CRLB values of the 15 human brain metabolites measured in the parietal and pgACC regions of normal brains at 7 T. The table demonstrates only the metabolites of the mean CRLBs below 20% except for Gln with a CRLB of 22.25 ± 1.2% in the parietal position. In addition, data from eight healthy volunteers have been analyzed using the LC-model. Therefore, there are two values of each metabolite for 3 each volunteer; one refers to the absolute concentration (in mM, density of brain 1050 kg/m ) and the other refers to the CRLB value (in %). The arithmetic mean and standard deviation of all metabolite concentrations and their CRLB values for 8àvolunteers have been calculated. For example, the concentration of Cr in the parietal region (mean ±SD) is 4.67 ± 0.93 mM for eight volunteers. Moreover, the CRLB of Cr (mean ±SD) is 2.25 ± 0.43% i.e., the Cr metabolite can be detected with high accuracy in all eight volunteers, with the CRLBs less than 3%. This result agrees with the data of a corresponding study at 7 T [14]. A comparison between the quantitative concentrations of glutamine, glutamate, N-acetylaspartate, and gammaaminobutyric at 3 T [28] and the concentrations at 3 and 7 T from this work in the pgACC region is shown in Table 2. The metabolite concentrations are given relative to that of creatine, as published in the corresponding study. The results of our measurements at 7 T agree well with the data of that study obtained using J-resolved PRESS 123
TABLE 1. Concentration Levels of Human Brain Metabolites of Eight Healthy Volunteers in the Parietal and pgACC Regions VOI = 3.38 ml (parietal region)
VOI = 3.5 ml (pgACC region)
Metabolites
Concentrations, mM Mean ± SD
CRLB, % Mean ± SD
Asp Cr GABA Gln Glu GPC GSH Ins NAA NAAG PCr PE Tau GPC + Cho Ins + Gly
2.42 ± 0.25 4.67 ± 0.93 1.23 ± 0.24 1.17 ± 0.31 6.45 ± 1.19 1.31 ± 0.29 1.38 ± 0.19 2.58 ± 0.36 10.82 ± 1.54 0.90 ± 0.27 2.91 ± 0.64 2.46 ± 0.89 1.2 ± 0.35 2.75 ± 0.43 3.28 ± 0.31
17.00 ± 4.42 2.25 ± 0.43 17.50 ± 6.02 22.25 ± 1.92 5.50 ± 1.12 7.25 ± 1.48 12.00 ± 0.03 7.25 ± 1.48 2.00 ± 0.00 7.75 ± 1.30 9.00 ± 2.00 13.00 ± 6.00 19.00 ± 8.00 3.1 ± 0.4 6.75 ± 1.64
Concentrations, mM Mean ± SD 2.08 ± 5.44 ± 1.36 ± 1.59 ± 7.74 ± 1.46 ± 0.74 ± 2.98 ± 8.51 ± 0.23± 1.79 ± 1.44 ± 1.32 ± 1.72 ± 3.95 ±
0.98 0.85 0.06 0.10 0.22 0.16 0.16 0.86 0.15 0.08 0.18 0.10 0.32 0.16 1.37
CRLB, % Mean ± SD 16.67 ± 8.08 2.83 ± 0.37 16.50 ± 3.25 12.08 ± 10.40 5.00 ± 1.15 6.50 ± 2.22 13.00 ± 3.00 7.00 ± 2.16 2.83 ± 0.13 16.58 ± 2.99 5.00 ± 1.80 11.00 ± 1.00 13.00 ± 3.26 2.67 ± 0.47 6.83 ± 1.57
TABLE 2. Comparison between the Relative Concentration Levels of Metabolites in pgACC Using STEAM with VERSE at 7 T and PRESS at 3 T (3 subjects) Relative to JPRESS at 3T [28]
Metabolites Cr (mM) GABA/Cr Gln/Cr Glu/Cr NAA/Cr
Measurements at 7 T (STEAM with VERSE)
Measurements at 3 T (PRESS)
Literature [28] at 3 T (JPRESS)
Mean ± SD
Mean ± SD
Mean ± SD
5.45 ± 0.85 0.25 ± 0.07 0.29 ± 0.12 1.42 ± 0.26 1.56 ± 0.17
7.51 ± 0.77 0.18 ± 0.13 0.25 ± 0.43 1.03 ± 0.63 1.53 ± 0.35
6.30 0.21 0.34 1.35 1.50
± ± ± ± ±
0.96 0.07 0.15 0.13 0.15
(JPRESS) spectroscopy with rather long scan times of 16 min and a large voxel size of approx. 17 ml compared to 1.75 ml in our 7 T data. A further advantage is that, even though JPRESS provides encoding of the J-coupling, thereby improving specificity in detecting J-coupled metabolites and alleviating a major limitation of 1D MRS [29], the use of STEAM adapted with VERSE at 7àT is sufficient to detect J-coupled human brain metabolites such as Gln, Glu, and GABA in shorter scan times. 1 Finally, the study shows an in vivo localization H MRS method employing STEAM with VERSE that allows SVS spectroscopy at 7 T using standard volume brain coils. With this method, it is possible to reliably measure fifteen metabolites in deep brain areas such as ACC. Two aspects contribute to the merits of the study: the small voxel size yields good homogeneity at 7 T leading to the narrow peaks, and the 24-channel coil may offer superior sensitivity compared to that of the 8-channel coil used at 3 T. 1 Conclusion. This study shows an in vivo H MRS method that combines STEAM with VERSE to reduce the RF peak power requirement and allows the use of the volume coils to acquire spectra from the deep brain structures. When corrected for the voxel volume and averages, the 7 T spectra show SNR improvement by a factor 1.94 com124
1
pared to those at 3 T. The results provide a H MRS-based non-invasive tool to detect and quantify 15 human brain metabolites in small volumes in deep brain structures at a high magnetic field of 7 T. Acknowledgments. The authors thank Dr. K. Zhong and O. Khorkhordin for their help with the adaptation of STEAM with VERSE. We also thank Dr. J. Kaufmann for his assistance in the preparation of the basis set and M. Walter for his help regarding neuroanatomy. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
G. Hasler, J. W. van der Veen, T. Tumonis, N. Meyers, J. Shen, and W. C. Drevets, Arch. Gen. Psychiatry, 64, 193–200 (2007). P. K. Mandal, Concept Magn. Res., 30, 40–64 (2007). P. A. Bottomley, Ann. N.Y. Acad. Sci., 508, 333–348 (1987). J. Frahm, K. D. Merboldt, and W. Haenicke, J. Magn. Reson., 72, 502–508 (1987). J. Kim, K. Chang, I. Song, S. Kim, B. Kwon, and M. Hee Han, Korean J. Radiol, 7, 156–161 (2006). C. Juchem, N. K. Logothetis, and J. Pfeuffer, Magn. Reson. Imaging, 25, 902–912 (2007). K. Ugurbil, G. Adriany, P. Andersen, W. Chen, M. Garwood, and R. Gruetter, Magn. Reson. Imaging, 21, 1263– 1281 (2003). P. B. Barker, D. O. Hearshen, and M. D. Boska, Magn. Reson. Med., 45, 765–769 (2001). S. Michaeli, M. Garwood, X. Zhu, L. DelaBarre, P. Andersen, G. Adriany, H. Merkle, K. Ugurbil, and W. Chen, Magn. Reson. Med., 47, 629–633 (2002). H. Qiao, X. Zhang, X. Zhu, F. Du, and W. Chen, Magn. Reson. Imaging, 24, 1281–1286 (2006). R. Gruetter, S. A. Weisdorf, V. Rajanayagan, M. Terpstra, H. Merkle, C. L. Truwit, M. Garwood, S. L. Nyberg, and K. Ugurbil, J. Magn. Reson., 135, 260–264 (1998). I. Tkac, P. Henry, P. Andersen, C. D. Keene, W. C. Low, and R. Gruetter, Magn. Reson. Med., 52 478–484 (2004). I. Tkac, R. Rao, M. K. Georgieff, and R. Gruetter, Magn. Reson. Med., 50, 24–32 (2003). R. Mekle, V. Mlyna’rik, G. Gambarota, M. Hergt, G. Krueger, and R. Gruetter, Magn. Reson. Med., 61, 1279– 1285 (2009). T. Michaelis, K. Merboldt, H. Bruhn, W. Haenicke, and J. Frahm, Neuroradiology, 187, 219–227 (1993). P. J. Pouwels, J. Frahm. MRM, 39, 53–60 (1998). P. B. Braker, B. J. Soher, S. J. Blackband, J. C. Chatham, V. P. Mathews, and R. N. Bryan, NMR Biomed., 6, 89–94 (1993). G. Jost, I. Harting, and S. Heiland, J. Magn. Reson. Imaging, 21, 66–71 (2005). G. Adriany and R. Gruetter, J. Magn. Reson., 125, 178–184 (1997). I. Tkac and R. Gruetter, Appl. Magn. Reson., 29, 139–157 (2005). B. A. Hargreaves, C. H. Cunningham, D. G. Nishimura, and S. M. Conolly, Magn. Reson. Med., 52, 590–597 (2004). M. Elywa, S. Mulla-Osman, M. Walter, K. Zhong, F. Godenschweger, O. Khorkhordin, J. Kaufmann, and O. Speck, Proc. Int. Soc. Magn. Reson. Med., 19 (2011). O. Khorkhordin, K. Zhong, F. Godenschweger, and O. Speck, Proc. Int. Soc. Magn. Reson. Med., 17 (2009). S. W. Provencher, Magn. Reson. Med., 30, 672–679 (1993). S. W. Provencher, NMR Biomed., 14, 260–264 (2001). M. Elywa, K. Zhong, and O. Speck, Arab. J. Nucl. Sci. Appl., 42, 131–140 (2009). M. Kanowski, J. Kaufmann, J. Braun, J. Bernarding, and C. Tempelmann, Magn. Reson. Med., 51, 904–912 (2004). M. Walter, A. Henning, S. Grimm, R. F. Schulte, J. Beck, U. Dydak, B. Schnepf, H. Boeker, P. Boesiger, and G. Northoff, Arch. Gen. Psychiatry, 66, 478–486 (2009). R. F. Schulte, T. Lange, J. Beck, D. Meier, P. and Boesiger, NMR Biomed., 19, 264–270 (2006).
125