Comparisons of Reproducibility and Mean Values of Diffusion Tensor Imaging-Derived Indices between Unipolar and Bipolar Diffusion Pulse Sequences Hsiao-Chien Miao, Ming-Ting Wu, E-Fong Kao, Yu-Hsien Chiu, Ming-Chung Chou From the Department of Medical Imaging and Radiological Sciences, Kaohsiung Medical University, Kaohsiung, Taiwan (H-CM, E-FK, M-CC); Department of Radiology, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan (M-TW); and Department of Healthcare Administration and Medical Informatics, Kaohsiung Medical University, Kaohsiung, Taiwan (Y-HC, M-CC).
ABSTRACT Eddy current distortion is an important issue that may influence the quantitative measurements of diffusion tensor imaging (DTI). The corrections of eddy current artifacts could be performed using bipolar diffusion gradients or unipolar gradients with affine registration. Whether the diffusion pulse sequence affects the quantification of DTI indices and the technique that produces more reliable DTI indices in terms of reproducibility both remain unclear. Therefore, the purpose of this study was to compare the reproducibility and mean values of DTI-derived indices between unipolar and bipolar diffusion pulse sequences based on actual human brain data. Five repeated datasets of unipolar and bipolar DTI were acquired from 10 healthy subjects at different echo times (TEs). The reproducibility and mean values of DTI indices were assessed by calculating the coefficient of variation and mean values of the 5 repeated measurements. The results revealed that the reproducibility and mean values of DTI indices were significantly affected by the pulse sequence. Unipolar DTI exhibited significantly higher reproducibility than bipolar DTI even at the same TE, and the mean values of DTI indices were significantly different between them. Therefore, we concluded that the reproducibility and mean values of DTI indices were significantly influenced by diffusion pulse sequences. Keywords: Unipolar, bipolar, DTI, reproducibility, mean value. Acceptance: Received August 29, 2014. Accepted for publication January 28, 2015. Correspondence: Address correspondence to Ming-Chung Chou, Department of Medical Imaging and Radiological Sciences, College of Health Sciences, Kaohsiung Medical University, 100, Shih-Chuan 1st Road, Kaohsiung, 80708, Taiwan. E-mail: [email protected]. J Neuroimaging 2015;25:892-899. DOI: 10.1111/jon.12231
Introduction Diffusion tensor imaging (DTI) has been widely utilized to characterize tissue architecture of the central nervous system by measuring 3-dimensional water diffusion in vivo.1 Many DTI indices were developed to quantify the diffusion property of the tissue, such as mean diffusivity (MD), fractional anisotropy (FA), and principal eigenvector (PEV), which were frequently harnessed to investigate the integrity and architecture of neuronal fibers.2–7 Diffusion measurements were found to be affected by a diffusion-weighting factor (or b-value),8–11 echo time (TE),10,12 signal-to-noise ratio (SNR),13,14 voxel size,15 and number of diffusion directions.16–18 In addition, eddy current distortions induced by a strong diffusion gradient also had profound effects on DTI measurements, leading to erroneous estimations of DTI indices, especially near the boundaries of brain regions.19–21 Therefore, previous studies performed DTI acquisition with bipolar diffusion gradients,19 twice-refocused spin echo diffusion,20 and the opposite diffusion gradient polarity method21 to effectively reduce eddy current distortions. However, TE and scan time were concomitantly lengthened in those techniques. In addition, affine image registration was also demonstrated to successfully reduce eddy current distortions in the unipolar DTI introduced by Stejkal and Tanner.22 Because the unipolar technique can achieve shorter TE under the same b-value, it potentially has a higher imaging SNR than other techniques.
892
Copyright
A previous study compared the correction of eddy current distortions between unipolar and bipolar DTI;23 however, the reproducibility and mean values of DTI indices have not previously been compared between the 2 techniques. Besides, the technique that produces more reliable DTI indices in terms of reproducibility remains unknown. Therefore, the purpose of this study was to employ a voxel-wise post-processing scheme, which is free of operating errors, to investigate the reproducibility and mean values of DTI indices between unipolar and bipolar DTI techniques based on in vivo human brain data.
Materials and Methods Data Acquisition Ten healthy volunteers (M:F = 5:5, aged 22.1 ± 2.7 years) who had no history of brain disease were enrolled in this study. The study was approved by the Local Institutional Review Board and an informed consent form was obtained from each participant prior to the MRI scan. All imaging data were acquired using a 3.0T MR scanner (Skyra, Erlangen, Siemens) with body coil excitation and a 20-channel phased array head coil for signal reception. DTI acquisition was performed using unipolar and bipolar diffusion gradients. The diagrams of 2 diffusion pulse sequences are shown in Figure 1. To compare their differences on DTI measurements, data were acquired at 3 TEs (71, 84, and 100 ms) for unipolar DTI and 2 TEs (84 and 100 ms) for bipolar DTI, where TEs of 71 and 84 ms were minimally
◦ 2015 by the American Society of Neuroimaging C
Fig 1. The pulse sequence diagrams for unipolar (A) and bipolar (B) techniques, where ࢞ is diffusion time and δ is gradient duration. Note that the eddy current induced by the diffusion gradients was eliminated in bipolar DTI.
available TEs in unipolar and bipolar DTIs, respectively. In unipolar DTI, the diffusion times for TEs of 71, 84, and 100 ms were around 28.6, 33.1, and 40.3 ms, respectively. In bipolar DTI, the diffusion times for TEs of 84 and 100 ms were around 12.9 and 15.3 ms, respectively. The scan was repeated 5 times to test reproducibility. Other imaging parameters for unipolar and bipolar pulse sequences were kept identical as follows: b-value = 1,000 s/mm2 TR = 2,700 ms, GRAPPA factor = 2.0, number of slice = 20, slice thickness = 2.5 mm, matrix size = 96 × 96, and FOV = 240 × 240 mm2 (isotropic resolution). In addition, the diffusion gradients were applied along 30
non-colinear directions plus one b0 image. The total scan time for both unipolar and bipolar DTI datasets was about 1 h and 10 min.
Data Processing All DTI datasets were transferred to a stand-alone workstation and processed offline using FSL (FMRIB Software Library, Oxford, UK). To correct the eddy current distortions in both unipolar and bipolar DTIs, 30 diffusion-weighted images (DWI) were first co-registered to the b0 image using FLIRT (FMRIB’s Linear Image Registration Tool) 12-parameter affine
Miao et al: Comparisons of Unipolar and Bipolar DTI
893
Fig 2. The RADAR post-processing scheme implemented in this study. After b0 images of all datasets were co-registered with each other, the original DWIs (A) were first co-registered with the mean b0 image (B) to obtain the 1st corrected DWIs (C). Subsequently, all of the 1st corrected DWIs were averaged to obtain the 1st average DWI (D), which was utilized to obtain the 2nd corrected DWIs (E) and the 2nd average DWI (F). Finally, the 2nd corrected DWIs were further co-registered with the 2nd average DWI to produce the 3rd corrected DWIs (G) for DTI calculation.
registration with a cost function of mutual information.24 Subsequently, the inter-scan rigid body motions of all DTI datasets were minimized using the RADAR (Robust Automated DWI Adaptive Registration) approach for both unipolar and bipolar datasets.25 The RADAR procedure used in this study is shown in Figure 2. The rotation matrices obtained from the previous step were utilized to adjust the gradient table accordingly to compensate for the rotational movements between scans.26 A brain mask generated by FSL BET (Brain Extraction Tool) was utilized to remove scalp signal and background noise.27 Finally, the diffusion tensor was normalized on a voxel-by-voxel basis to obtain 3 eigenvalues (λ1 , λ2 , and λ3 ) and 3 corresponding eigenvectors (e1 , e2 , and e3 ), from which MD and FA maps were calculated using the following 2 equations: λ1 + λ2 + λ3 , and 3 3 (λ1 − M D)2 + (λ2 − M D)2 + (λ3 − M D)2 FA = . 2 λ21 + λ22 + λ23
MD =
The first eigenvector e1 , which is in the direction with the largest eigenvalue, was defined as PEV to assess the fiber orientations. To provide a benchmark for comparing DTI indices between different DTI datasets, all of DTI datasets after eddy current distortion correction were concatenated without being averaged to serve as the reference DTI dataset. The reference DTI indices, including FAref and MDref , obtained from the reference DTI dataset were further harnessed to separate gray and white matter tissue. In this study, the gray and white matter 894
regions were segmented based on the referenced DTI dataset using the following criteria: FAref > .25 and MDref < .001 mm2 /s for white matter and FAref ࣚ .25 and MDref < .001 mm2 /s for gray matter, according to the results of previous DTI studies on normal aging subjects.28–30 Representative segmentation results are shown in Figure 3.
Statistical Analysis This study assessed the reproducibility of DTI indices by calculating the coefficient of variation (CV) on a voxel-by-voxel basis measured in 5 repeated DTI datasets.10,14 The reproducibility of PEV orientation was assessed by estimating the angular variance (AV) measured in 5 repeated datasets.10,14,17 The angular variance of PEV in N scans was given by the following equation:
AV(PEV) =
N 1 cos−1 (PEVi · PEV), N i=1
where PEVi is the individual PEV and PEV is the normalized vector sum of all PEV components. Because the DTI indices were demonstrated to be dependent on imaging parameters, it was particularly difficult to define the actual gold standard DTI dataset to assess accuracy. Therefore, this study only compared the mean values of DTI indices between unipolar and bipolar DTI to see if there exist significant differences of DTI indices
Journal of Neuroimaging Vol 25 No 6 November/December 2015
The reproducibility and mean values of DTI indices in gray and white matter regions of all subjects were statistically compared between unipolar and bipolar DTI using paired Student’s t-test. In addition, repeated-measure ANOVA was performed to test if the reproducibility and mean values varied with TE. The difference was considered significant if P < .05.
Results
Fig 3. The gray (A) and white (B) matter regions obtained based on the reference FA (C) and MD (D) maps.
between them. The mean values of DTI indices were calculated from 5 DTI measurements using the following 2 equations: FA=
5 i=1
F Ai , and M D =
5 i=1
M Di .
The results of CV(FA), CV(MD), and AV(PEV) showed that unipolar DTI has significantly higher reproducibility of DTI indices than bipolar DTI at different TEs, as shown in Figure 4, and the variations were found to generally increase with TE. The statistical comparisons demonstrated that unipolar DTI had significantly higher reproducibility of DTI indices than bipolar DTI in both gray and white matter at different TEs; however, CV(MD) was not significantly different between the 2 techniques at TE = 100 ms, as shown in Figure 5. Specifically, unipolar DTI exhibited significantly lower CV(FA) than bipolar DTI at minimum TE (71 vs. 84 ms) and also at TE = 84 ms and 100 ms in both gray and white matter. Similarly, unipolar DTI exhibited significantly lower CV(MD) than bipolar DTI at minimum TE (71 vs. 84 ms) and at TE = 84 ms in both gray and white matter. Although CV(MD) was found to be significantly different at TE = 100 ms in white matter, it did not reach a significant difference in gray matter between the 2 techniques. In PEV, unipolar DTI showed significantly lower AV than bipolar DTI at TE = 84 and 100 ms in both gray and white matters, and unipolar DTI also exhibited significantly lower AV than that of bipolar DTI at minimum TE (71 vs. 84 ms) in both gray and white matter. The means and R1-9 standard deviations of CV(FA), CV(MD), and AV(PEV) obtained from 10 healthy subjects are listed in Table 1.
Fig 4. The CV(FA) (A), CV(MD) (B), and AV(PEV) (C) maps obtained from one representative subject.
Miao et al: Comparisons of Unipolar and Bipolar DTI
895
Fig 5. The CV(FA) (A), CV(MD) (B), and AV(PEV) (C) of gray (GM) and white matter (WM) tissue obtained from 10 healthy subjects. The asterisks (*) indicate a significant difference. The FA and MD values of unipolar and bipolar DTI were roughly consistent at different TEs in both gray and white matter tissue, but the statistical analysis revealed significant difference between them, as shown in Figure 6. The statistical results showed that FA and MD values were significantly different between unipolar and bipolar DTI at TE = 84 and 100 ms in both gray and white matter. The FA and MD values were also significantly different between the 2 techniques at minimum TE (71 ms vs. 84 ms), and bipolar DTI showed significantly higher FA and MD values than unipolar DTI at the same TE (84 ms and 100 ms). The means and standard deviations of FA and MD values obtained from 10 healthy subjects are listed in Table 2. The repeated-measure analysis revealed that in unipolar DTI, CV(FA) varied significantly with TE in gray matter, whereas CV(MD) varied significantly with TE in white matter. AV(PEV) varied significantly with TE in both gray and white matter tissue in unipolar DTI. However, CV(FA), CV(MD), and AV(PEV) in bipolar DTI did not significantly vary with
896
TE in both gray and white matter tissue. In both unipolar and bipolar DTI, FA and MD values were found to significantly vary with TE in both gray and white matter tissue.
Discussion This study compared the reproducibility and mean values of DTI indices between unipolar and bipolar DTI. The results indicated that the variations of DTI indices increased with TE in both unipolar and bipolar DTI datasets, suggesting that shorter TE increased the imaging SNR so as to improve the image quality. Because unipolar DTI could reach a TE shorter than that of bipolar DTI, the unipolar DTI is potentially superior to bipolar DTI when a higher SNR is required. However, at identical TE, the results revealed that unipolar DTI exhibited significantly lower variations of DTI indices than bipolar DTI, which might be attributable to the difference of diffusion time between the 2 techniques. A previous simulation study has shown that the
Journal of Neuroimaging Vol 25 No 6 November/December 2015
Table 1. Means and Standard Deviations of CV(FA), CV(MD), and AV(PEV) Obtained from 10 Healthy Subjects CV(FA)% WM TE (ms)
Unipolar
CV(MD)% Bipolar
AV(PEV) Degree
Unipolar
Bipolar
Unipolar
Bipolar
71 84 100
7.2 ± 2.1ξ 8.0 ± 2.8* 7.2 ± 1.2**
8.9 ± 2.7*ξ 8.4 ± 1.4**
4.5 ± .4ξ 5.2 ± .6* 4.8 ± .6**
5.9 ± .7*ξ 5.7 ± 1.0**
5.4 ± .6ξ 6.0 ± .8* 6.1 ± .9**
6.9 ± .8*ξ 7.2 ± 1.1**
GM TE (ms) 71 84 100
14.9 ± 2.3ξ 16.2 ± 3.0* 16.0 ± 1.7**
17.8 ± 17.8 ± 1.5**
4.6 ± .3ξ 5.1 ± .4* 5.0 ± .8**
5.4 ± 5.4 ± .8**
15.4 ± 1.3ξ 17.4 ± 1.6* 18.0 ± 2.2**
20.0 ± 1.6*ξ 21.1 ± 2.3**
2.5*ξ
.3*ξ
The symbols (ξ , *, **) indicate significant difference between unipolar and bipolar DTI at minimum TE (ξ ), TE = 84 (*), and TE = 100 (**) ms.
Fig 6. The FA (A) and MD (B) values of gray (GM) and white matter (WM) tissue obtained from 10 healthy subjects. The asterisks (*) indicate a significant difference.
transverse diffusion of white matter decreases with diffusion times shorter than 15 ms, but it tends to be constant when diffusion time exceeds 30 ms.31 However, the unrestricted longitudinal diffusion would increase slightly with diffusion time in the direction parallel to fiber orientations. Because unipolar DTI had a longer diffusion time than bipolar DTI, the water molecules tended to be more likely to interact with the boundary of the cell membrane in the unipolar DTI and hence led to more stable DTI indices than bipolar DTI at an identical TE. The results of repeated-measure analysis indicated that TE significantly affected the reproducibility of DTI indices in unipolar DTI. Because a longer TE led to a lower imaging SNR, CV(FA), CV(MD), and AV(PEV) were slightly increased with TE in both gray and white matter tissue. However in bipolar
DTI, the repeated-measure analysis did not reveal significant changes of CV(FA), CV(MD), and AV(PEV) with TE, suggesting that the decreased imaging SNR did not significantly alter the reproducibility of DTI indices at TE from 84 ms to 100 ms. This study also revealed that FA and MD values of white matter were significantly different between unipolar and bipolar DTI techniques at different TEs, which was likely owing to the inherent difference of diffusion time between them. The longer the TE was, the larger the differences of both FA and MD values would be between the 2 techniques. Such a phenomenon might be attributable to the increased difference in diffusion time between the 2 techniques at longer TE. In repeated-measure analysis, the FA and MD values of both unipolar and bipolar techniques were found to vary significantly with TE in both Miao et al: Comparisons of Unipolar and Bipolar DTI
897
Table 2. Means and Standard Deviations of FA and MD Obtained from 10 Healthy Subjects MD x10−3 mm2 /s
FA WM TE (ms)
Unipolar
Bipolar
Unipolar
.012ξ
Bipolar
.016ξ
71 84 100
.427 ± .430 ± .009* .409 ± .011**
.436 ± .011*ξ .418 ± .008**
.732 ± .735 ± .016* .738 ± .018**
.744 ± .017*ξ .754 ± .019**
GM TE (ms) 71 84 100
.141 ± .005 .137 ± .004* .127 ± .006**
.141 ± .006* .132 ± .006**
.798 ± .036ξ .808 ± .039* .827 ± .040**
.819 ± .044*ξ .847 ± .045**
The symbols (ξ , *, **) indicate significant difference between unipolar and bipolar DTI at minimum TE (ξ ), TE = 84 (*), and TE = 100 (**) ms.
gray and white matter tissue. The increased MD values with TE in both unipolar and bipolar DTI techniques may be due to the fact that longer TE led to longer diffusion time so that wider displacement of water molecules could be achieved. In contrast, FA values were found to decrease with TE in gray matter in both unipolar and bipolar techniques, suggesting that longer diffusion time would result in less differences of displacement of water molecules in different diffusion directions in the isotropic region. However, the results also showed that longer diffusion time did not lead to a monotonic increase of FA in white matter. Because longer TE also gave rise to a lower imaging SNR, and hence lower FA in white matter,14 the results found in this study were likely affected simultaneously by both diffusion time and imaging SNR as TE was changed. Moreover, it has been shown that the eddy current distortions in unipolar DTI may not be fully corrected by 12-parameter affine registration.22 A previous study also demonstrated that there existed residual eddy current distortions in bipolar DTI,23 and a magnetic field monitoring approach could be used to minimize the distortions. However, such an approach necessitated a field camera comprising multiple NMR probes to measure eddy current phases and may not be suitable for clinical applications. Hence, to minimize the eddy current distortions, this study performed 12-parameter affine registration in both unipolar and bipolar DTI data. However, the residual errors in unipolar and bipolar DTI data may contribute to additional differences between them. Nevertheless, the results of this study showed that unipolar DTI has significantly higher reproducibility compared with bipolar DTI, suggesting that unipolar DTI is potentially superior to bipolar DTI and would be a more suitable technique for clinical applications when higher SNR and longer diffusion time are needed.
Conclusions In conclusion, this study compared the reproducibility and mean values of DTI indices between unipolar and bipolar DTI at 3.0 Tesla. The results showed that the reproducibility and mean values of DTI indices were significantly different between the 2 techniques. Therefore, we concluded that the reproducibility and accuracy of DTI indices are significantly impacted by diffusion pulse sequences. This study was supported in part by grant KMU-Q110001 from Kaohsiung Medical University Research Foundation.
898
References 1. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259-67. 2. Pajevic S, Pierpaoli C. Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: application to white matter fiber tract mapping in the human brain. Magn Reson Med 1999;42:526-40. 3. Mori S, Crain BJ, Chacko VP, et al. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol 1999;45:265-9. 4. Douek P, Turner R, Pekar J, et al. MR color mapping of myelin fiber orientation. J Comput Assist Tomogr 1991;15:923-9. 5. Chien D, Buxton RB, Kwong KK, et al. MR diffusion imaging of the human brain. J Comput Assist Tomo 1990;14:514-20. 6. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209-19. 7. Basser PJ, Pajevic S, Pierpaoli C, et al. In vivo fiber tractography using DT-MRI data. Magn Reson Med 2000;44:625-32. 8. Pereira RS, Harris AD, Sevick RJ, et al. Effect of b value on contrast during diffusion-weighted magnetic resonance imaging assessment of acute ischemic stroke. J Magn Reson Imaging 2002;15: 591-6. 9. Hui ES, Cheung MM, Chan KC, et al. B-value dependence of DTI quantitation and sensitivity in detecting neural tissue changes. Neuroimage 2010;49:2366-74. 10. Chou MC, Kao EF, Mori S. Effects of b-value and echo time on magnetic resonance diffusion tensor imaging-derived parameters at 1.5 T: a voxel-wise study. J Med Biol Eng 2013;33:45-50. 11. Melhem ER, Itoh R, Jones L, et al. Diffusion tensor MR imaging of the brain: effect of diffusion weighting on trace and anisotropy measurements. Am J Neuroradiol 2000;21:1813-20. 12. Qin W, Yu CS, Zhang F, et al. Effects of echo time on diffusion quantification of brain white matter at 1.5T and 3.0T. Magn Reson Med 2009;61:755-60. 13. Saritas EU, Lee JH, Nishimura DG. SNR dependence of optimal parameters for apparent diffusion coefficient measurements. IEEE T Med Imaging 2011;30:424-37. 14. Farrell JAD, Landman BA, Jones CK, et al. Effects of signal-tonoise ratio on the accuracy and reproducibility, of diffusion tensor imaging-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5T. J Magn Reson Imaging 2007;26:756-67. 15. Oouchi H, Yamada K, Sakai K, et al. Diffusion anisotropy measurement of brain white matter is affected by voxel size: underestimation occurs in areas with crossing fibers. Am J Neuroradiol 2007;28:1102-6. 16. Ni H, Kavcic V, Zhu T, et al. Effects of number of diffusion gradient directions on derived diffusion tensor imaging indices in human brain. Am J Neuroradiol 2006;27:1776-81. 17. Landman BA, Farrell JAD, Jones CK, et al. Effects of diffusion weighting schemes on the reproducibility of DTI-derived fractional
Journal of Neuroimaging Vol 25 No 6 November/December 2015
18.
19.
20.
21.
22.
23.
anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5T. Neuroimage 2007;36:1123-38. Giannelli M, Cosottini M, Michelassi MC, et al. Dependence of brain DTI maps of fractional anisotropy and mean diffusivity on the number of diffusion weighting directions. J Appl Clin Med Phys 2010;11:176-90. Alexander AL, Tsuruda JS, Parker DL. Elimination of eddy current artifacts in diffusion-weighted echo-planar images: the use of bipolar gradients. Magn Reson Med 1997;38:1016-21. Reese TG, Heid O, Weisskoff RM, et al. Reduction of eddy-currentinduced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson Med 2003;49:177-82. Bodammer N, Kaufmann J, Kanowski M, et al. Eddy current correction in diffusion-weighted imaging using pairs of images acquired with opposite diffusion gradient polarity. Magn Reson Med 2004;51:188-93. Mohammadi S, Moller HE, Kugel H, et al. Correcting eddy current and motion effects by affine whole-brain registrations: evaluation of three-dimensional distortions and comparison with slicewise correction. Magn Reson Med 2010;64:1047-56. Chan RW, von Deuster C, Giese D, et al. Characterization and correction of eddy-current artifacts in unipolar and bipolar diffusion sequences using magnetic field monitoring. J Magn Reson 2014;244:74-84.
24. Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5:14356. 25. Landman BA, Farrell JAD, Mori S, et al. On the coregistration of diffusion weighted images. In: Book of abstracts: Fourteenth International Society for Magnetic Resonance in Medicine, Seattle, WA: ISMRM; 2006: p. 2987. 26. Leemans A, Jones DK. The B-matrix must be rotated when correcting for subject motion in DTI data. Magn Reson Med 2009;61:1336-49. 27. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp 2002;17:143-55. 28. Gong NJ, Wong CS, Chan CC, et al. Aging in deep gray matter and white matter revealed by diffusional kurtosis imaging. Neurobiol Aging 2014;35:2203-16. 29. Sullivan EV, Adalsteinsson E, Hedehus M, et al. Equivalent disruption of regional white matter microstructure in ageing healthy men and women. Neuroreport 2001;12:99– 104. 30. Wang Q, Xu X, Zhang M. Normal aging in the basal ganglia evaluated by eigenvalues of diffusion tensor imaging. Am J Neuroradiol 2010;31:516-20. 31. Szafer A, Zhong JH, Gore JC. Theoretical-model for water diffusion in tissues. Magn Reson Med 1995;33:697-712.
Miao et al: Comparisons of Unipolar and Bipolar DTI
899
ABSTRACT Eddy current distortion is an important issue that may influence the quantitative measurements of diffusion tensor imaging (DTI). The corrections of eddy current artifacts could be performed using bipolar diffusion gradients or unipolar gradients with affine registration. Whether the diffusion pulse sequence affects the quantification of DTI indices and the technique that produces more reliable DTI indices in terms of reproducibility both remain unclear. Therefore, the purpose of this study was to compare the reproducibility and mean values of DTI-derived indices between unipolar and bipolar diffusion pulse sequences based on actual human brain data. Five repeated datasets of unipolar and bipolar DTI were acquired from 10 healthy subjects at different echo times (TEs). The reproducibility and mean values of DTI indices were assessed by calculating the coefficient of variation and mean values of the 5 repeated measurements. The results revealed that the reproducibility and mean values of DTI indices were significantly affected by the pulse sequence. Unipolar DTI exhibited significantly higher reproducibility than bipolar DTI even at the same TE, and the mean values of DTI indices were significantly different between them. Therefore, we concluded that the reproducibility and mean values of DTI indices were significantly influenced by diffusion pulse sequences. Keywords: Unipolar, bipolar, DTI, reproducibility, mean value. Acceptance: Received August 29, 2014. Accepted for publication January 28, 2015. Correspondence: Address correspondence to Ming-Chung Chou, Department of Medical Imaging and Radiological Sciences, College of Health Sciences, Kaohsiung Medical University, 100, Shih-Chuan 1st Road, Kaohsiung, 80708, Taiwan. E-mail: [email protected]. J Neuroimaging 2015;25:892-899. DOI: 10.1111/jon.12231
Introduction Diffusion tensor imaging (DTI) has been widely utilized to characterize tissue architecture of the central nervous system by measuring 3-dimensional water diffusion in vivo.1 Many DTI indices were developed to quantify the diffusion property of the tissue, such as mean diffusivity (MD), fractional anisotropy (FA), and principal eigenvector (PEV), which were frequently harnessed to investigate the integrity and architecture of neuronal fibers.2–7 Diffusion measurements were found to be affected by a diffusion-weighting factor (or b-value),8–11 echo time (TE),10,12 signal-to-noise ratio (SNR),13,14 voxel size,15 and number of diffusion directions.16–18 In addition, eddy current distortions induced by a strong diffusion gradient also had profound effects on DTI measurements, leading to erroneous estimations of DTI indices, especially near the boundaries of brain regions.19–21 Therefore, previous studies performed DTI acquisition with bipolar diffusion gradients,19 twice-refocused spin echo diffusion,20 and the opposite diffusion gradient polarity method21 to effectively reduce eddy current distortions. However, TE and scan time were concomitantly lengthened in those techniques. In addition, affine image registration was also demonstrated to successfully reduce eddy current distortions in the unipolar DTI introduced by Stejkal and Tanner.22 Because the unipolar technique can achieve shorter TE under the same b-value, it potentially has a higher imaging SNR than other techniques.
892
Copyright
A previous study compared the correction of eddy current distortions between unipolar and bipolar DTI;23 however, the reproducibility and mean values of DTI indices have not previously been compared between the 2 techniques. Besides, the technique that produces more reliable DTI indices in terms of reproducibility remains unknown. Therefore, the purpose of this study was to employ a voxel-wise post-processing scheme, which is free of operating errors, to investigate the reproducibility and mean values of DTI indices between unipolar and bipolar DTI techniques based on in vivo human brain data.
Materials and Methods Data Acquisition Ten healthy volunteers (M:F = 5:5, aged 22.1 ± 2.7 years) who had no history of brain disease were enrolled in this study. The study was approved by the Local Institutional Review Board and an informed consent form was obtained from each participant prior to the MRI scan. All imaging data were acquired using a 3.0T MR scanner (Skyra, Erlangen, Siemens) with body coil excitation and a 20-channel phased array head coil for signal reception. DTI acquisition was performed using unipolar and bipolar diffusion gradients. The diagrams of 2 diffusion pulse sequences are shown in Figure 1. To compare their differences on DTI measurements, data were acquired at 3 TEs (71, 84, and 100 ms) for unipolar DTI and 2 TEs (84 and 100 ms) for bipolar DTI, where TEs of 71 and 84 ms were minimally
◦ 2015 by the American Society of Neuroimaging C
Fig 1. The pulse sequence diagrams for unipolar (A) and bipolar (B) techniques, where ࢞ is diffusion time and δ is gradient duration. Note that the eddy current induced by the diffusion gradients was eliminated in bipolar DTI.
available TEs in unipolar and bipolar DTIs, respectively. In unipolar DTI, the diffusion times for TEs of 71, 84, and 100 ms were around 28.6, 33.1, and 40.3 ms, respectively. In bipolar DTI, the diffusion times for TEs of 84 and 100 ms were around 12.9 and 15.3 ms, respectively. The scan was repeated 5 times to test reproducibility. Other imaging parameters for unipolar and bipolar pulse sequences were kept identical as follows: b-value = 1,000 s/mm2 TR = 2,700 ms, GRAPPA factor = 2.0, number of slice = 20, slice thickness = 2.5 mm, matrix size = 96 × 96, and FOV = 240 × 240 mm2 (isotropic resolution). In addition, the diffusion gradients were applied along 30
non-colinear directions plus one b0 image. The total scan time for both unipolar and bipolar DTI datasets was about 1 h and 10 min.
Data Processing All DTI datasets were transferred to a stand-alone workstation and processed offline using FSL (FMRIB Software Library, Oxford, UK). To correct the eddy current distortions in both unipolar and bipolar DTIs, 30 diffusion-weighted images (DWI) were first co-registered to the b0 image using FLIRT (FMRIB’s Linear Image Registration Tool) 12-parameter affine
Miao et al: Comparisons of Unipolar and Bipolar DTI
893
Fig 2. The RADAR post-processing scheme implemented in this study. After b0 images of all datasets were co-registered with each other, the original DWIs (A) were first co-registered with the mean b0 image (B) to obtain the 1st corrected DWIs (C). Subsequently, all of the 1st corrected DWIs were averaged to obtain the 1st average DWI (D), which was utilized to obtain the 2nd corrected DWIs (E) and the 2nd average DWI (F). Finally, the 2nd corrected DWIs were further co-registered with the 2nd average DWI to produce the 3rd corrected DWIs (G) for DTI calculation.
registration with a cost function of mutual information.24 Subsequently, the inter-scan rigid body motions of all DTI datasets were minimized using the RADAR (Robust Automated DWI Adaptive Registration) approach for both unipolar and bipolar datasets.25 The RADAR procedure used in this study is shown in Figure 2. The rotation matrices obtained from the previous step were utilized to adjust the gradient table accordingly to compensate for the rotational movements between scans.26 A brain mask generated by FSL BET (Brain Extraction Tool) was utilized to remove scalp signal and background noise.27 Finally, the diffusion tensor was normalized on a voxel-by-voxel basis to obtain 3 eigenvalues (λ1 , λ2 , and λ3 ) and 3 corresponding eigenvectors (e1 , e2 , and e3 ), from which MD and FA maps were calculated using the following 2 equations: λ1 + λ2 + λ3 , and 3 3 (λ1 − M D)2 + (λ2 − M D)2 + (λ3 − M D)2 FA = . 2 λ21 + λ22 + λ23
MD =
The first eigenvector e1 , which is in the direction with the largest eigenvalue, was defined as PEV to assess the fiber orientations. To provide a benchmark for comparing DTI indices between different DTI datasets, all of DTI datasets after eddy current distortion correction were concatenated without being averaged to serve as the reference DTI dataset. The reference DTI indices, including FAref and MDref , obtained from the reference DTI dataset were further harnessed to separate gray and white matter tissue. In this study, the gray and white matter 894
regions were segmented based on the referenced DTI dataset using the following criteria: FAref > .25 and MDref < .001 mm2 /s for white matter and FAref ࣚ .25 and MDref < .001 mm2 /s for gray matter, according to the results of previous DTI studies on normal aging subjects.28–30 Representative segmentation results are shown in Figure 3.
Statistical Analysis This study assessed the reproducibility of DTI indices by calculating the coefficient of variation (CV) on a voxel-by-voxel basis measured in 5 repeated DTI datasets.10,14 The reproducibility of PEV orientation was assessed by estimating the angular variance (AV) measured in 5 repeated datasets.10,14,17 The angular variance of PEV in N scans was given by the following equation:
AV(PEV) =
N 1 cos−1 (PEVi · PEV), N i=1
where PEVi is the individual PEV and PEV is the normalized vector sum of all PEV components. Because the DTI indices were demonstrated to be dependent on imaging parameters, it was particularly difficult to define the actual gold standard DTI dataset to assess accuracy. Therefore, this study only compared the mean values of DTI indices between unipolar and bipolar DTI to see if there exist significant differences of DTI indices
Journal of Neuroimaging Vol 25 No 6 November/December 2015
The reproducibility and mean values of DTI indices in gray and white matter regions of all subjects were statistically compared between unipolar and bipolar DTI using paired Student’s t-test. In addition, repeated-measure ANOVA was performed to test if the reproducibility and mean values varied with TE. The difference was considered significant if P < .05.
Results
Fig 3. The gray (A) and white (B) matter regions obtained based on the reference FA (C) and MD (D) maps.
between them. The mean values of DTI indices were calculated from 5 DTI measurements using the following 2 equations: FA=
5 i=1
F Ai , and M D =
5 i=1
M Di .
The results of CV(FA), CV(MD), and AV(PEV) showed that unipolar DTI has significantly higher reproducibility of DTI indices than bipolar DTI at different TEs, as shown in Figure 4, and the variations were found to generally increase with TE. The statistical comparisons demonstrated that unipolar DTI had significantly higher reproducibility of DTI indices than bipolar DTI in both gray and white matter at different TEs; however, CV(MD) was not significantly different between the 2 techniques at TE = 100 ms, as shown in Figure 5. Specifically, unipolar DTI exhibited significantly lower CV(FA) than bipolar DTI at minimum TE (71 vs. 84 ms) and also at TE = 84 ms and 100 ms in both gray and white matter. Similarly, unipolar DTI exhibited significantly lower CV(MD) than bipolar DTI at minimum TE (71 vs. 84 ms) and at TE = 84 ms in both gray and white matter. Although CV(MD) was found to be significantly different at TE = 100 ms in white matter, it did not reach a significant difference in gray matter between the 2 techniques. In PEV, unipolar DTI showed significantly lower AV than bipolar DTI at TE = 84 and 100 ms in both gray and white matters, and unipolar DTI also exhibited significantly lower AV than that of bipolar DTI at minimum TE (71 vs. 84 ms) in both gray and white matter. The means and R1-9 standard deviations of CV(FA), CV(MD), and AV(PEV) obtained from 10 healthy subjects are listed in Table 1.
Fig 4. The CV(FA) (A), CV(MD) (B), and AV(PEV) (C) maps obtained from one representative subject.
Miao et al: Comparisons of Unipolar and Bipolar DTI
895
Fig 5. The CV(FA) (A), CV(MD) (B), and AV(PEV) (C) of gray (GM) and white matter (WM) tissue obtained from 10 healthy subjects. The asterisks (*) indicate a significant difference. The FA and MD values of unipolar and bipolar DTI were roughly consistent at different TEs in both gray and white matter tissue, but the statistical analysis revealed significant difference between them, as shown in Figure 6. The statistical results showed that FA and MD values were significantly different between unipolar and bipolar DTI at TE = 84 and 100 ms in both gray and white matter. The FA and MD values were also significantly different between the 2 techniques at minimum TE (71 ms vs. 84 ms), and bipolar DTI showed significantly higher FA and MD values than unipolar DTI at the same TE (84 ms and 100 ms). The means and standard deviations of FA and MD values obtained from 10 healthy subjects are listed in Table 2. The repeated-measure analysis revealed that in unipolar DTI, CV(FA) varied significantly with TE in gray matter, whereas CV(MD) varied significantly with TE in white matter. AV(PEV) varied significantly with TE in both gray and white matter tissue in unipolar DTI. However, CV(FA), CV(MD), and AV(PEV) in bipolar DTI did not significantly vary with
896
TE in both gray and white matter tissue. In both unipolar and bipolar DTI, FA and MD values were found to significantly vary with TE in both gray and white matter tissue.
Discussion This study compared the reproducibility and mean values of DTI indices between unipolar and bipolar DTI. The results indicated that the variations of DTI indices increased with TE in both unipolar and bipolar DTI datasets, suggesting that shorter TE increased the imaging SNR so as to improve the image quality. Because unipolar DTI could reach a TE shorter than that of bipolar DTI, the unipolar DTI is potentially superior to bipolar DTI when a higher SNR is required. However, at identical TE, the results revealed that unipolar DTI exhibited significantly lower variations of DTI indices than bipolar DTI, which might be attributable to the difference of diffusion time between the 2 techniques. A previous simulation study has shown that the
Journal of Neuroimaging Vol 25 No 6 November/December 2015
Table 1. Means and Standard Deviations of CV(FA), CV(MD), and AV(PEV) Obtained from 10 Healthy Subjects CV(FA)% WM TE (ms)
Unipolar
CV(MD)% Bipolar
AV(PEV) Degree
Unipolar
Bipolar
Unipolar
Bipolar
71 84 100
7.2 ± 2.1ξ 8.0 ± 2.8* 7.2 ± 1.2**
8.9 ± 2.7*ξ 8.4 ± 1.4**
4.5 ± .4ξ 5.2 ± .6* 4.8 ± .6**
5.9 ± .7*ξ 5.7 ± 1.0**
5.4 ± .6ξ 6.0 ± .8* 6.1 ± .9**
6.9 ± .8*ξ 7.2 ± 1.1**
GM TE (ms) 71 84 100
14.9 ± 2.3ξ 16.2 ± 3.0* 16.0 ± 1.7**
17.8 ± 17.8 ± 1.5**
4.6 ± .3ξ 5.1 ± .4* 5.0 ± .8**
5.4 ± 5.4 ± .8**
15.4 ± 1.3ξ 17.4 ± 1.6* 18.0 ± 2.2**
20.0 ± 1.6*ξ 21.1 ± 2.3**
2.5*ξ
.3*ξ
The symbols (ξ , *, **) indicate significant difference between unipolar and bipolar DTI at minimum TE (ξ ), TE = 84 (*), and TE = 100 (**) ms.
Fig 6. The FA (A) and MD (B) values of gray (GM) and white matter (WM) tissue obtained from 10 healthy subjects. The asterisks (*) indicate a significant difference.
transverse diffusion of white matter decreases with diffusion times shorter than 15 ms, but it tends to be constant when diffusion time exceeds 30 ms.31 However, the unrestricted longitudinal diffusion would increase slightly with diffusion time in the direction parallel to fiber orientations. Because unipolar DTI had a longer diffusion time than bipolar DTI, the water molecules tended to be more likely to interact with the boundary of the cell membrane in the unipolar DTI and hence led to more stable DTI indices than bipolar DTI at an identical TE. The results of repeated-measure analysis indicated that TE significantly affected the reproducibility of DTI indices in unipolar DTI. Because a longer TE led to a lower imaging SNR, CV(FA), CV(MD), and AV(PEV) were slightly increased with TE in both gray and white matter tissue. However in bipolar
DTI, the repeated-measure analysis did not reveal significant changes of CV(FA), CV(MD), and AV(PEV) with TE, suggesting that the decreased imaging SNR did not significantly alter the reproducibility of DTI indices at TE from 84 ms to 100 ms. This study also revealed that FA and MD values of white matter were significantly different between unipolar and bipolar DTI techniques at different TEs, which was likely owing to the inherent difference of diffusion time between them. The longer the TE was, the larger the differences of both FA and MD values would be between the 2 techniques. Such a phenomenon might be attributable to the increased difference in diffusion time between the 2 techniques at longer TE. In repeated-measure analysis, the FA and MD values of both unipolar and bipolar techniques were found to vary significantly with TE in both Miao et al: Comparisons of Unipolar and Bipolar DTI
897
Table 2. Means and Standard Deviations of FA and MD Obtained from 10 Healthy Subjects MD x10−3 mm2 /s
FA WM TE (ms)
Unipolar
Bipolar
Unipolar
.012ξ
Bipolar
.016ξ
71 84 100
.427 ± .430 ± .009* .409 ± .011**
.436 ± .011*ξ .418 ± .008**
.732 ± .735 ± .016* .738 ± .018**
.744 ± .017*ξ .754 ± .019**
GM TE (ms) 71 84 100
.141 ± .005 .137 ± .004* .127 ± .006**
.141 ± .006* .132 ± .006**
.798 ± .036ξ .808 ± .039* .827 ± .040**
.819 ± .044*ξ .847 ± .045**
The symbols (ξ , *, **) indicate significant difference between unipolar and bipolar DTI at minimum TE (ξ ), TE = 84 (*), and TE = 100 (**) ms.
gray and white matter tissue. The increased MD values with TE in both unipolar and bipolar DTI techniques may be due to the fact that longer TE led to longer diffusion time so that wider displacement of water molecules could be achieved. In contrast, FA values were found to decrease with TE in gray matter in both unipolar and bipolar techniques, suggesting that longer diffusion time would result in less differences of displacement of water molecules in different diffusion directions in the isotropic region. However, the results also showed that longer diffusion time did not lead to a monotonic increase of FA in white matter. Because longer TE also gave rise to a lower imaging SNR, and hence lower FA in white matter,14 the results found in this study were likely affected simultaneously by both diffusion time and imaging SNR as TE was changed. Moreover, it has been shown that the eddy current distortions in unipolar DTI may not be fully corrected by 12-parameter affine registration.22 A previous study also demonstrated that there existed residual eddy current distortions in bipolar DTI,23 and a magnetic field monitoring approach could be used to minimize the distortions. However, such an approach necessitated a field camera comprising multiple NMR probes to measure eddy current phases and may not be suitable for clinical applications. Hence, to minimize the eddy current distortions, this study performed 12-parameter affine registration in both unipolar and bipolar DTI data. However, the residual errors in unipolar and bipolar DTI data may contribute to additional differences between them. Nevertheless, the results of this study showed that unipolar DTI has significantly higher reproducibility compared with bipolar DTI, suggesting that unipolar DTI is potentially superior to bipolar DTI and would be a more suitable technique for clinical applications when higher SNR and longer diffusion time are needed.
Conclusions In conclusion, this study compared the reproducibility and mean values of DTI indices between unipolar and bipolar DTI at 3.0 Tesla. The results showed that the reproducibility and mean values of DTI indices were significantly different between the 2 techniques. Therefore, we concluded that the reproducibility and accuracy of DTI indices are significantly impacted by diffusion pulse sequences. This study was supported in part by grant KMU-Q110001 from Kaohsiung Medical University Research Foundation.
898
References 1. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994;66:259-67. 2. Pajevic S, Pierpaoli C. Color schemes to represent the orientation of anisotropic tissues from diffusion tensor data: application to white matter fiber tract mapping in the human brain. Magn Reson Med 1999;42:526-40. 3. Mori S, Crain BJ, Chacko VP, et al. Three-dimensional tracking of axonal projections in the brain by magnetic resonance imaging. Ann Neurol 1999;45:265-9. 4. Douek P, Turner R, Pekar J, et al. MR color mapping of myelin fiber orientation. J Comput Assist Tomogr 1991;15:923-9. 5. Chien D, Buxton RB, Kwong KK, et al. MR diffusion imaging of the human brain. J Comput Assist Tomo 1990;14:514-20. 6. Basser PJ, Pierpaoli C. Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. J Magn Reson B 1996;111:209-19. 7. Basser PJ, Pajevic S, Pierpaoli C, et al. In vivo fiber tractography using DT-MRI data. Magn Reson Med 2000;44:625-32. 8. Pereira RS, Harris AD, Sevick RJ, et al. Effect of b value on contrast during diffusion-weighted magnetic resonance imaging assessment of acute ischemic stroke. J Magn Reson Imaging 2002;15: 591-6. 9. Hui ES, Cheung MM, Chan KC, et al. B-value dependence of DTI quantitation and sensitivity in detecting neural tissue changes. Neuroimage 2010;49:2366-74. 10. Chou MC, Kao EF, Mori S. Effects of b-value and echo time on magnetic resonance diffusion tensor imaging-derived parameters at 1.5 T: a voxel-wise study. J Med Biol Eng 2013;33:45-50. 11. Melhem ER, Itoh R, Jones L, et al. Diffusion tensor MR imaging of the brain: effect of diffusion weighting on trace and anisotropy measurements. Am J Neuroradiol 2000;21:1813-20. 12. Qin W, Yu CS, Zhang F, et al. Effects of echo time on diffusion quantification of brain white matter at 1.5T and 3.0T. Magn Reson Med 2009;61:755-60. 13. Saritas EU, Lee JH, Nishimura DG. SNR dependence of optimal parameters for apparent diffusion coefficient measurements. IEEE T Med Imaging 2011;30:424-37. 14. Farrell JAD, Landman BA, Jones CK, et al. Effects of signal-tonoise ratio on the accuracy and reproducibility, of diffusion tensor imaging-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5T. J Magn Reson Imaging 2007;26:756-67. 15. Oouchi H, Yamada K, Sakai K, et al. Diffusion anisotropy measurement of brain white matter is affected by voxel size: underestimation occurs in areas with crossing fibers. Am J Neuroradiol 2007;28:1102-6. 16. Ni H, Kavcic V, Zhu T, et al. Effects of number of diffusion gradient directions on derived diffusion tensor imaging indices in human brain. Am J Neuroradiol 2006;27:1776-81. 17. Landman BA, Farrell JAD, Jones CK, et al. Effects of diffusion weighting schemes on the reproducibility of DTI-derived fractional
Journal of Neuroimaging Vol 25 No 6 November/December 2015
18.
19.
20.
21.
22.
23.
anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5T. Neuroimage 2007;36:1123-38. Giannelli M, Cosottini M, Michelassi MC, et al. Dependence of brain DTI maps of fractional anisotropy and mean diffusivity on the number of diffusion weighting directions. J Appl Clin Med Phys 2010;11:176-90. Alexander AL, Tsuruda JS, Parker DL. Elimination of eddy current artifacts in diffusion-weighted echo-planar images: the use of bipolar gradients. Magn Reson Med 1997;38:1016-21. Reese TG, Heid O, Weisskoff RM, et al. Reduction of eddy-currentinduced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson Med 2003;49:177-82. Bodammer N, Kaufmann J, Kanowski M, et al. Eddy current correction in diffusion-weighted imaging using pairs of images acquired with opposite diffusion gradient polarity. Magn Reson Med 2004;51:188-93. Mohammadi S, Moller HE, Kugel H, et al. Correcting eddy current and motion effects by affine whole-brain registrations: evaluation of three-dimensional distortions and comparison with slicewise correction. Magn Reson Med 2010;64:1047-56. Chan RW, von Deuster C, Giese D, et al. Characterization and correction of eddy-current artifacts in unipolar and bipolar diffusion sequences using magnetic field monitoring. J Magn Reson 2014;244:74-84.
24. Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5:14356. 25. Landman BA, Farrell JAD, Mori S, et al. On the coregistration of diffusion weighted images. In: Book of abstracts: Fourteenth International Society for Magnetic Resonance in Medicine, Seattle, WA: ISMRM; 2006: p. 2987. 26. Leemans A, Jones DK. The B-matrix must be rotated when correcting for subject motion in DTI data. Magn Reson Med 2009;61:1336-49. 27. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp 2002;17:143-55. 28. Gong NJ, Wong CS, Chan CC, et al. Aging in deep gray matter and white matter revealed by diffusional kurtosis imaging. Neurobiol Aging 2014;35:2203-16. 29. Sullivan EV, Adalsteinsson E, Hedehus M, et al. Equivalent disruption of regional white matter microstructure in ageing healthy men and women. Neuroreport 2001;12:99– 104. 30. Wang Q, Xu X, Zhang M. Normal aging in the basal ganglia evaluated by eigenvalues of diffusion tensor imaging. Am J Neuroradiol 2010;31:516-20. 31. Szafer A, Zhong JH, Gore JC. Theoretical-model for water diffusion in tissues. Magn Reson Med 1995;33:697-712.
Miao et al: Comparisons of Unipolar and Bipolar DTI
899
