Magnetic Resonance Imaging 32 (2014) 625–629
Contents lists available at ScienceDirect
Magnetic Resonance Imaging journal homepage: www.mrijournal.com
Multiple sclerosis: Benefits of q-space imaging in evaluation of normal-appearing and periplaque white matter Masaaki Hori a,⁎, Mariko Yoshida a, Kazumasa Yokoyama b, Koji Kamagata a, Fumitaka Kumagai a, c, Issei Fukunaga a, c, Kouhei Kamiya a, Michimasa Suzuki a, Yoshitaka Masutani d, Nozomi Hamasaki a, Yuriko Suzuki a, e, Shinsuke Kyogoku f, Nobutaka Hattori b, Shigeki Aoki a a
Department of Radiology, Juntendo University School of Medicine, Tokyo, Japan Department of Neurology, Juntendo University School of Medicine, Tokyo, Japan c Department of Health Science, Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo, Japan d Department of Radiology, The University of Tokyo Hospital, Tokyo, Japan e Philips Electronics Japan, Tokyo, Japan f Department of Radiology, Juntendo University Urayasu Hospital, Chiba, Japan b
a r t i c l e
i n f o
Article history: Received 6 July 2013 Revised 12 February 2014 Accepted 14 February 2014 Keywords: Multiple sclerosis Diffusion tensor imaging q-space imaging Root mean square displacement Periplaque white matter Normal-appearing white matter
a b s t r a c t Introduction: Diffusion tensor imaging (DTI) reveals white matter pathology in patients with multiple sclerosis (MS). A recent non-Gaussian diffusion imaging technique, q-space imaging (QSI), may provide several advantages over conventional MRI techniques in regard to in vivo evaluation of the disease process in patients with MS. The purpose of this study is to investigate the use of root mean square displacement (RMSD) derived from QSI data to characterize plaques, periplaque white matter (PWM), and normalappearing white matter (NAWM) in patients with MS. Methods: We generated apparent diffusion coefficient (ADC) and fractional anisotropy (FA) maps by using conventional DTI data from 21 MS patients; we generated RMSD maps by using QSI data from these patients. We used the Steel–Dwass test to compare the diffusion metrics of regions of interest in plaques, PWM, and NAWM. Results: ADC differed (P b 0.05) between plaques and PWM and between plaques and NAWM. FA differed (P b 0.05) between plaques and NAWM. RMSD differed (P b 0.05) between plaques and PWM, plaques and NAWM, and PWM and NAWM. Conclusion: RMSD values from QSI may reflect microstructural changes and white-matter damage in patients with MS with higher sensitivity than do conventional ADC and FA values. © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).
1. Introduction Several imaging techniques are potentially useful for elucidating the disease process in patients with multiple sclerosis (MS). In addition to conventional MRI techniques (including T2-weighted imaging), quantitative brain MRI techniques such as diffusionweighted imaging (DWI) and its derivative technique, diffusion tensor imaging (DTI), enable MS lesions to be characterized in vivo according to quantitative values, such as fractional anisotropy (FA) and the apparent diffusion coefficient (ADC). In addition, DWI and DTI offer advantages over conventional techniques in their ability to ⁎ Corresponding author at: Department of Radiology, School of Medicine, Juntendo University, 2-1-1 Hongo, Bunkyo-ku, Tokyo, 113-8421, Japan. Tel.: +81 3 3813 3111; fax: +81 3 3816 0958. E-mail address: [email protected] (M. Hori).
detect otherwise hidden abnormalities in normal-appearing white matter (NAWM) [1–5]. Moreover, DTI has been reported to reveal differences in white matter abnormality between the white matter at the periphery of plaques and distant NAWM [1]. Non-Gaussian diffusion MRI techniques, including q-space imaging (QSI) analysis [6–8] and diffusional kurtosis imaging (DKI) [9], have emerged recently. Unlike DWI and DTI, QSI and DKI do not require the assumption of a Gaussian shape when modeling the distribution of free water molecules. QSI and DKI have yielded promising results in the evaluation of brain [10–13] and spinal cord [14–18] disorders in vivo because they provide diffusion metrics, such as the root mean square displacement (RMSD), that are additional to, and different from, those of Gaussian techniques. In addition, DKI has demonstrated its usefulness in characterizing the disease process in patients with MS [6,19,20]. In particular, RMSD values obtained from QSI data reflect the full extent of water molecule movement and provide more accurate
http://dx.doi.org/10.1016/j.mri.2014.02.024 © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).
626
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
microstructural information than do ADC and FA values [8,21]. We therefore hypothesized that RMSD values derived from QSI analysis would provide more information on in vivo structural and pathologic changes in the brains of patients with MS, and at higher sensitivity, than do conventional DTI metrics. Our aim here was to investigate the use of RMSD derived from QSI data to characterize plaques, periplaque white matter (PWM), and NAWM in patients with MS. 2. Materials and methods 2.1. Patients Between December 2011 and August 2012, we evaluated a total of 21 consecutive patients with relapsing–remitting (n = 20) or secondary progressive (n = 1) MS (6 male; 15 female; age [mean ± 1 SD], 44.3 ± 10.06 years; median [range] Expanded Disability Status Scale score [22], 2.0 [0.0–6.0]) who had a previously established diagnosis of MS according to 2005 revisions to the McDonald Criteria [23] without acute plaques. Informed consent was obtained from each patient. We obtained ethics approval from the institutional review board before the study.
described [8,24]. Briefly, the key principle in q-space analysis is that a Fourier transform of the signal attenuation with regard to q provides the PDF for diffusion by using multiple q-values [17]. The shape of the computed PDF can be characterized by the FWHM and the maximum height of the curve. In the condition of unrestricted Gaussian diffusion, the diffusion constant D and the RMSD for one-dimensional diffusion can be computed from the FWHM. Mean RMSD was calculated from the FWHM values (RMSD = 0.425 × FWHM) [16,17]. By referring to conventional MR images, two experienced neuroradiologists (M.Y. and M.H.) manually placed ovoid region of interests (ROIs) on b = 0 QSI data by using dTV II FZR and VolumeOne 1.81 software (Image Computing and Analysis Laboratory, Department of Radiology, The University of Tokyo Hospital). ROIs were drawn in plaques (defined as areas of abnormally high signal intensity on the b = 0 q-space image), periplaque white matter (PWM; defined as a white-matter area that had normal signal intensity and was closest to a plaque), and NAWM (defined as an area of WM with normal signal intensity that was contralateral to a plaque; Fig. 1) [1]. The dTV II FZR software allowed for copying of the ROIs and guaranteed the evaluation of the same region with diffusion metric maps. The average FA, ADC, and FWHM values in each ROI were measured; areas with severe signal loss or calculation errors were excluded from analysis.
2.2. Image acquisition All images were acquired on a 3-T scanner (Achieva, Philips Medical Systems, Best, The Netherlands). After routine MRI comprising turbo spin-echo T2-weighted and fluid-attenuated inversion-recovery axial imaging, we acquired T1-weighted, sagittal 3D magnetizationprepared rapid-acquisition gradient-echo and QSI data. Imaging parameters for conventional axial images were: repetition time (ms)/ echo time (ms): 4000/100 for T2-weighted imaging, 10000/100 for fluid-attenuated inversion-recovery axial imaging; number of signals acquired, two; section thickness/gap, 5/1 mm; 22 sections; and pixel size, 0.45 × 0.45 mm. Imaging parameters for magnetization-prepared rapid-acquisition gradient-echo imaging were: repetition time (ms)/echo time (ms), 15/3.5; number of signals acquired, one; section thickness/gap, 0.86/0 mm; 170 sections; and pixel size, 0.81 × 0.81 mm. Parameters used for QSI were: repetition time (ms)/echo time (ms), 4000/96; number of signals acquired, one; section thickness/gap, 4/0 mm; 10 sections; field of view, 256 × 256 mm; matrix, 64 × 64; imaging time, 4 min 36 s; and 12 b-values (0, 124, 496, 1116, 1983, 3099, 4463, 6074, 7934, 10041, 12397 and 15000 s/mm2), with diffusion encoding in 6 directions for every b-value. The q-value was linearly incremented from 0 to 104.64 cm−1 [16,19,24]. The gradient length (δ) and time between the two leading edges of the diffusion gradient (Δ) were 37.8 and 47.3 ms, respectively. QSI was limited to large, semioval areas of white matter to minimize the scanning time to that appropriate for clinical use.
2.4. Statistical analysis The three areas (plaques, PWM, and NAWM) were compared according to the Steel–Dwass test for multiple comparisons by using the statistical software package R (Version 2.8.1). A P value of less than 0.05 was considered to indicate a statistically significant difference. Interrater reliability was assessed by using Pearson’s correlation coefficient. 3. Results Data from all 22 patients were included in the evaluation, without fatal image degeneration or artifacts. Fig. 2 shows representative b = 0 DTI image (echo-planar T2-weighted image), FA, and ADC maps generated by using conventional DTI data, and an RMSD map created from QSI data. All plaques yielded low values on FA maps and high values on both RMSD and ADC maps.
2.3. Analysis of DTI and QSI data After we corrected for distortions due to eddy currents using an affine registration on the magnetic resonance imager, diffusion tensor and q-space analyses were performed with dTV II FZR and Volume-One 1.81 software (Image Computing and Analysis Laboratory, Department of Radiology, The University of Tokyo Hospital, Tokyo, Japan) [8] on a stand-alone personal computer running Windows (Microsoft, Redmond, WA, USA). ADC and FA were calculated pixel-by-pixel according to the conventional mono-exponential model from part of the q-space data, b-values of 0 and 1116 s/mm 2, because these data included multiple b-value data. Next, the full width at half maximum (FWHM) of the probability density function (PDF) was calculated as previously
Fig. 1. ROIs manually placed in the plaque (region 1), PWM (region 2), and NAWM (region 3) on a b = 0 diffusion-weighted image of a patient with MS.
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
627
Fig. 2. Another patient’s representative (A) b = 0 DTI (echo-planar T2-weighted image), (B) ADC, and (C) FA maps from conventional DTI data, and (D) RMSD map from QSI data. Corresponding T2- and T1-weighted images (E, F) at the same slice level for visual comparisons.
628
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
Reproducibility was expressed in terms of the interrater correlation coefficient; the coefficient was 0.86 for the ADC analysis, 0.79 for the FA analysis, and 0.94 for the RMSD analysis. ADC values (mean ± 1 SD) for plaques, PWM, and NAWM were 0.640 ± 0.116, 0.545 ± 0.091, 0.490 ± 0.043 (10 − 3 mm 2 /s), respectively. FA values for plaques, PWM, and NAWM were 0.271 ± 0.072, 0.298 ± 0.080, 0.355 ± 0.092. RMSD values for plaques, PWM, and NAWM were 5.805 ± 1.201, 4.981 ± 0.857, 4.435 ± 0.400 μm, respectively. ADC values differed between plaques and PWM (P b 0.001) and between plaques and NAWM (P b 0.001). FA differed significantly (P b 0.001) between plaques and NAWM. RMSD data differed between plaques and PWM (P = 0.038), between plaques and NAWM (P b 0.001), and between PWM and NAWM (P = 0.019).
The third limitation of our study was the small number of patients evaluated and the lack of clinical correlations with diffusional metrics. FA and ADC values of the white matter can be influenced by duration and severity of MS. Therefore, before the usefulness of RMSD as an imaging biomarker can be established, longitudinal studies and correlations between RMSD and clinical disease characteristics must be established. In conclusion, RMSD values derived from QSI data may reflect microstructural changes and damage in the white matter of patients with MS with higher sensitivity than do ADC and FA values obtained from conventional DTI. More studies of the imaging–pathology relationship are needed, but QSI has the potential to provide new information for characterizing MS pathology in vivo. Conflict of interest statement
4. Discussion Our findings of highest ADC values and lowest FA values in plaques followed by PWM and NAWM are consistent with those of previous studies [1,24], and these patterns can be explained in part by the severity of white matter damage. In addition, RMSD values decreased from plaques to PWM and then NAWM; these changes varied significantly depending on the distance from the plaque. In a previous report addressing correlations between brain pathology and findings on imaging, the authors concluded that slight increases in ADC may be indicative of axonal loss, and decreases in FA may signal microglial activation in the white matter without plaques [25]. Our results showed that only RMSD was significantly different among plaques, PWM, and NAWM. Therefore, compared with conventional diffusion metrics, RMSD values from QSI may be a more sensitive biomarker to detect such graded pathologic change in white matter. The precise reason for the high sensitivity of RMSD in this regard remains unknown as yet. One explanation may lie in the fact that QSI uses multiple b-value data including high-b values (over 10000 s/mm2), which indicate intracellular water components, whereas conventional DTI is believed to measure water molecules in the extracellular space [6]. Moreover, QSI is a non-Gaussian diffusion analysis, with which it is possible (at least theoretically) to measure the full extent of water-molecule movement without having to assume Gaussian distribution of data, unlike the situation for conventional DTI. Therefore, QSI and its metric RMSD can lead to better estimation of actual neural tissue microstructural changes in vivo. One potential limitation of our study is the limited coverage obtained of the brain through QSI scanning (4 mm × 10 slices) and the relatively poor spatial resolution of 4-mm isovoxels. We used this condition to reduce the scan time to a clinically feasible duration. However, future investigations should focus on increasing both brain coverage and spatial resolution. Currently available techniques are limited in their ability to decrease scanning time on the MR scanners available in the clinical setting. However, various advanced techniques, such as compressed sensing [26], are expected to overcome this problem. Moreover, inherently lower SNR was expected in the calculated FA and ADC maps because they were calculated using data of only two b values and 6 motion probing gradient (MPG) axes and may substantially affect the results. Although we recognize the value of a greater number of MPG directions for more precise estimation of FA and ADC, we used the lower number because of the limited scan time for clinical practice. Another limitation is that QSI-derived P0 (probability for zero displacement) map was not used for the analysis in this study. We recognized that the P0 map was useful for MS lesion detection [6,19,27]. However, we thought that it was difficult to use P0 values for quantitative analysis because the values of P0 were usually scaled as arbitrary unit.
The authors declare that there are no conflicts of interest. Acknowledgments We thank Shuji Sato for help with data acquisition. This study was supported by a Grant-in-Aid for Scientific Research on Innovative Areas (Comprehensive Brain Science Network) from the Ministry of Education, Science, Sports, and Culture of Japan. This work was supported by JSPS KAKENHI Grant Number 24591788. References [1] Guo AC, MacFall JR, Provenzale JM. Multiple sclerosis: diffusion tensor MR imaging for evaluation of normal-appearing white matter. Radiology 2002;222:729–36. [2] Rovaris M, Bozzali M, Iannucci G, Ghezzi A, Caputo D, Montanari E, et al. Assessment of normal-appearing white and gray matter in patients with primary progressive multiple sclerosis: a diffusion-tensor magnetic resonance imaging study. Arch Neurol 2002;59:1406–12. [3] Rocca MA, Iannucci G, Rovaris M, Comi G, Filippi M. Occult tissue damage in patients with primary progressive multiple sclerosis is independent of T2-visible lesions–a diffusion tensor MR study. J Neurol 2003;250:456–60. [4] Inglese M, Bester M. Diffusion imaging in multiple sclerosis: research and clinical implications. NMR Biomed 2010;23:865–72. [5] Filippi M, Rocca MA, De Stefano N, Enzinger C, Fisher E, Horsfield MA, et al. Magnetic resonance techniques in multiple sclerosis: the present and the future. Arch Neurol 2011;68:1514–20. [6] Assaf Y, Ben-Bashat D, Chapman J, Peled S, Biton IE, Kafri M, et al. High b-value q-space analyzed diffusion-weighted MRI: application to multiple sclerosis. Magn Reson Med 2002;47:115–26. [7] Cohen Y, Assaf Y. High b-value q-space analyzed diffusion-weighted MRS and MRI in neuronal tissues – a technical review. NMR Biomed 2002;15:516–42. [8] Hori M, Fukunaga I, Masutani Y, Taoka T, Kamagata K, Suzuki Y, et al. Visualizing non-Gaussian diffusion: clinical application of q-space imaging and diffusional kurtosis imaging of the brain and spine. Magn Reson Med Sci 2012;11:221–33. [9] Jensen JH, Helpern JA, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005;53:1432–40. [10] Hori M, Motosugi U, Fatima Z, Kumagai H, Ikenaga S, Ishigame K, et al. A Comparison of Mean Displacement Values Using High b-Value Q-Space Diffusion-weighted MRI with Conventional Apparent Diffusion Coefficients in Patients with Stroke. Acad Radiol 2011;18:837–41. [11] Fatima Z, Motosugi U, Waqar AB, Hori M, Ishigame K, Oishi N, et al. Associations among q-space MRI, diffusion-weighted MRI and histopathological parameters in meningiomas. Eur Radiol 2013;23:2258–63. [12] Yamada K, Sakai K, Akazawa K, Sugimoto N, Nakagawa M, Mizuno T. Detection of early neuronal damage in CADASIL patients by q-space MR imaging. Neuroradiology 2013;55:283–90. [13] Kafri M, Sasson E, Assaf Y, Balash Y, Aiznstein O, Hausdorff JM, et al. High-level gait disorder: associations with specific white matter changes observed on advanced diffusion imaging. J Neuroimaging 2013;23:39–46. [14] Hori M, Motosug U, Fatima Z, Ishigame K, Araki T. Mean displacement map of spine and spinal cord disorders using high b-value q-space imaging-feasibility study. Acta Radiol 2011;52:1155–8. [15] Hori M, Fukunaga I, Masutani Y, Nakanishi A, Shimoji K, Kamagata K, et al. New diffusion metrics for spondylotic myelopathy at an early clinical stage. Eur Radiol 2012;22:1797–802. [16] Farrell JA, Smith SA, Gordon-Lipkin EM, Reich DS, Calabresi PA, van Zijl PC. High b-value q-space diffusion-weighted MRI of the human cervical spinal cord in vivo: feasibility and application to multiple sclerosis. Magn Reson Med 2008;59:1079–89.
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629 [17] Farrell JA, Zhang J, Jones MV, Deboy CA, Hoffman PN, Landman BA, et al. q-space and conventional diffusion imaging of axon and myelin damage in the rat spinal cord after axotomy. Magn Reson Med 2010;63: 1323–35. [18] Raz E, Bester M, Sigmund EE, Tabesh A, Babb JS, Jaggi H, et al. A Better Characterization of Spinal Cord Damage in Multiple Sclerosis: A Diffusional Kurtosis Imaging Study. AJNR Am J Neuroradiol 2013;34:1846–52. [19] Assaf Y, Chapman J, Ben-Bashat D, Hendler T, Segev Y, Korczyn AD, et al. White matter changes in multiple sclerosis: correlation of q-space diffusion MRI and 1H MRS. Magn Reson Imaging 2005;23:703–10. [20] Yoshida M, Hori M, Yokoyama K, Fukunaga I, Suzuki M, Kamagata K, et al. Diffusional kurtosis imaging of normal-appearing white matter in multiple sclerosis: preliminary clinical experience. Jpn J Radiol 2013;31:50–5. [21] Fatima Z, Motosugi U, Hori M, Ishigame K, Kumagai H, Ikenaga S, et al. q-space imaging (QSI) of the brain: comparison of displacement parameters by QSI and DWI. Magn Reson Med Sci 2010;9:109–10.
629
[22] Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 1983;33:1444–52. [23] Polman CH, Reingold SC, Edan G, Filippi M, Hartung HP, Kappos L, et al. Diagnostic criteria for multiple sclerosis: 2005 revisions to the "McDonald Criteria". Ann Neurol 2005;58:840–6. [24] Assaf Y, Mayk A, Cohen Y. Displacement imaging of spinal cord using q-space diffusion-weighted MRI. Magn Reson Med 2000;44:713–22. [25] Moll NM, Rietsch AM, Thomas S, Ransohoff AJ, Lee JC, Fox R, et al. Multiple sclerosis normal-appearing white matter: pathology-imaging correlations. Ann Neurol 2011;70:764–73. [26] Menzel MI, Tan ET, Khare K, Sperl JI, King KF, Tao X, et al. Accelerated diffusion spectrum imaging in the human brain using compressed sensing. Magn Reson Med 2011;66:1226–33. [27] Pagani E, Bammer R, Horsfield MA, Rovaris M, Gass A, Ciccarelli O, et al. Diffusion MR imaging in multiple sclerosis: technical aspects and challenges. AJNR Am J Neuroradiol 2007;28:411–20.
Contents lists available at ScienceDirect
Magnetic Resonance Imaging journal homepage: www.mrijournal.com
Multiple sclerosis: Benefits of q-space imaging in evaluation of normal-appearing and periplaque white matter Masaaki Hori a,⁎, Mariko Yoshida a, Kazumasa Yokoyama b, Koji Kamagata a, Fumitaka Kumagai a, c, Issei Fukunaga a, c, Kouhei Kamiya a, Michimasa Suzuki a, Yoshitaka Masutani d, Nozomi Hamasaki a, Yuriko Suzuki a, e, Shinsuke Kyogoku f, Nobutaka Hattori b, Shigeki Aoki a a
Department of Radiology, Juntendo University School of Medicine, Tokyo, Japan Department of Neurology, Juntendo University School of Medicine, Tokyo, Japan c Department of Health Science, Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo, Japan d Department of Radiology, The University of Tokyo Hospital, Tokyo, Japan e Philips Electronics Japan, Tokyo, Japan f Department of Radiology, Juntendo University Urayasu Hospital, Chiba, Japan b
a r t i c l e
i n f o
Article history: Received 6 July 2013 Revised 12 February 2014 Accepted 14 February 2014 Keywords: Multiple sclerosis Diffusion tensor imaging q-space imaging Root mean square displacement Periplaque white matter Normal-appearing white matter
a b s t r a c t Introduction: Diffusion tensor imaging (DTI) reveals white matter pathology in patients with multiple sclerosis (MS). A recent non-Gaussian diffusion imaging technique, q-space imaging (QSI), may provide several advantages over conventional MRI techniques in regard to in vivo evaluation of the disease process in patients with MS. The purpose of this study is to investigate the use of root mean square displacement (RMSD) derived from QSI data to characterize plaques, periplaque white matter (PWM), and normalappearing white matter (NAWM) in patients with MS. Methods: We generated apparent diffusion coefficient (ADC) and fractional anisotropy (FA) maps by using conventional DTI data from 21 MS patients; we generated RMSD maps by using QSI data from these patients. We used the Steel–Dwass test to compare the diffusion metrics of regions of interest in plaques, PWM, and NAWM. Results: ADC differed (P b 0.05) between plaques and PWM and between plaques and NAWM. FA differed (P b 0.05) between plaques and NAWM. RMSD differed (P b 0.05) between plaques and PWM, plaques and NAWM, and PWM and NAWM. Conclusion: RMSD values from QSI may reflect microstructural changes and white-matter damage in patients with MS with higher sensitivity than do conventional ADC and FA values. © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).
1. Introduction Several imaging techniques are potentially useful for elucidating the disease process in patients with multiple sclerosis (MS). In addition to conventional MRI techniques (including T2-weighted imaging), quantitative brain MRI techniques such as diffusionweighted imaging (DWI) and its derivative technique, diffusion tensor imaging (DTI), enable MS lesions to be characterized in vivo according to quantitative values, such as fractional anisotropy (FA) and the apparent diffusion coefficient (ADC). In addition, DWI and DTI offer advantages over conventional techniques in their ability to ⁎ Corresponding author at: Department of Radiology, School of Medicine, Juntendo University, 2-1-1 Hongo, Bunkyo-ku, Tokyo, 113-8421, Japan. Tel.: +81 3 3813 3111; fax: +81 3 3816 0958. E-mail address: [email protected] (M. Hori).
detect otherwise hidden abnormalities in normal-appearing white matter (NAWM) [1–5]. Moreover, DTI has been reported to reveal differences in white matter abnormality between the white matter at the periphery of plaques and distant NAWM [1]. Non-Gaussian diffusion MRI techniques, including q-space imaging (QSI) analysis [6–8] and diffusional kurtosis imaging (DKI) [9], have emerged recently. Unlike DWI and DTI, QSI and DKI do not require the assumption of a Gaussian shape when modeling the distribution of free water molecules. QSI and DKI have yielded promising results in the evaluation of brain [10–13] and spinal cord [14–18] disorders in vivo because they provide diffusion metrics, such as the root mean square displacement (RMSD), that are additional to, and different from, those of Gaussian techniques. In addition, DKI has demonstrated its usefulness in characterizing the disease process in patients with MS [6,19,20]. In particular, RMSD values obtained from QSI data reflect the full extent of water molecule movement and provide more accurate
http://dx.doi.org/10.1016/j.mri.2014.02.024 © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).
626
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
microstructural information than do ADC and FA values [8,21]. We therefore hypothesized that RMSD values derived from QSI analysis would provide more information on in vivo structural and pathologic changes in the brains of patients with MS, and at higher sensitivity, than do conventional DTI metrics. Our aim here was to investigate the use of RMSD derived from QSI data to characterize plaques, periplaque white matter (PWM), and NAWM in patients with MS. 2. Materials and methods 2.1. Patients Between December 2011 and August 2012, we evaluated a total of 21 consecutive patients with relapsing–remitting (n = 20) or secondary progressive (n = 1) MS (6 male; 15 female; age [mean ± 1 SD], 44.3 ± 10.06 years; median [range] Expanded Disability Status Scale score [22], 2.0 [0.0–6.0]) who had a previously established diagnosis of MS according to 2005 revisions to the McDonald Criteria [23] without acute plaques. Informed consent was obtained from each patient. We obtained ethics approval from the institutional review board before the study.
described [8,24]. Briefly, the key principle in q-space analysis is that a Fourier transform of the signal attenuation with regard to q provides the PDF for diffusion by using multiple q-values [17]. The shape of the computed PDF can be characterized by the FWHM and the maximum height of the curve. In the condition of unrestricted Gaussian diffusion, the diffusion constant D and the RMSD for one-dimensional diffusion can be computed from the FWHM. Mean RMSD was calculated from the FWHM values (RMSD = 0.425 × FWHM) [16,17]. By referring to conventional MR images, two experienced neuroradiologists (M.Y. and M.H.) manually placed ovoid region of interests (ROIs) on b = 0 QSI data by using dTV II FZR and VolumeOne 1.81 software (Image Computing and Analysis Laboratory, Department of Radiology, The University of Tokyo Hospital). ROIs were drawn in plaques (defined as areas of abnormally high signal intensity on the b = 0 q-space image), periplaque white matter (PWM; defined as a white-matter area that had normal signal intensity and was closest to a plaque), and NAWM (defined as an area of WM with normal signal intensity that was contralateral to a plaque; Fig. 1) [1]. The dTV II FZR software allowed for copying of the ROIs and guaranteed the evaluation of the same region with diffusion metric maps. The average FA, ADC, and FWHM values in each ROI were measured; areas with severe signal loss or calculation errors were excluded from analysis.
2.2. Image acquisition All images were acquired on a 3-T scanner (Achieva, Philips Medical Systems, Best, The Netherlands). After routine MRI comprising turbo spin-echo T2-weighted and fluid-attenuated inversion-recovery axial imaging, we acquired T1-weighted, sagittal 3D magnetizationprepared rapid-acquisition gradient-echo and QSI data. Imaging parameters for conventional axial images were: repetition time (ms)/ echo time (ms): 4000/100 for T2-weighted imaging, 10000/100 for fluid-attenuated inversion-recovery axial imaging; number of signals acquired, two; section thickness/gap, 5/1 mm; 22 sections; and pixel size, 0.45 × 0.45 mm. Imaging parameters for magnetization-prepared rapid-acquisition gradient-echo imaging were: repetition time (ms)/echo time (ms), 15/3.5; number of signals acquired, one; section thickness/gap, 0.86/0 mm; 170 sections; and pixel size, 0.81 × 0.81 mm. Parameters used for QSI were: repetition time (ms)/echo time (ms), 4000/96; number of signals acquired, one; section thickness/gap, 4/0 mm; 10 sections; field of view, 256 × 256 mm; matrix, 64 × 64; imaging time, 4 min 36 s; and 12 b-values (0, 124, 496, 1116, 1983, 3099, 4463, 6074, 7934, 10041, 12397 and 15000 s/mm2), with diffusion encoding in 6 directions for every b-value. The q-value was linearly incremented from 0 to 104.64 cm−1 [16,19,24]. The gradient length (δ) and time between the two leading edges of the diffusion gradient (Δ) were 37.8 and 47.3 ms, respectively. QSI was limited to large, semioval areas of white matter to minimize the scanning time to that appropriate for clinical use.
2.4. Statistical analysis The three areas (plaques, PWM, and NAWM) were compared according to the Steel–Dwass test for multiple comparisons by using the statistical software package R (Version 2.8.1). A P value of less than 0.05 was considered to indicate a statistically significant difference. Interrater reliability was assessed by using Pearson’s correlation coefficient. 3. Results Data from all 22 patients were included in the evaluation, without fatal image degeneration or artifacts. Fig. 2 shows representative b = 0 DTI image (echo-planar T2-weighted image), FA, and ADC maps generated by using conventional DTI data, and an RMSD map created from QSI data. All plaques yielded low values on FA maps and high values on both RMSD and ADC maps.
2.3. Analysis of DTI and QSI data After we corrected for distortions due to eddy currents using an affine registration on the magnetic resonance imager, diffusion tensor and q-space analyses were performed with dTV II FZR and Volume-One 1.81 software (Image Computing and Analysis Laboratory, Department of Radiology, The University of Tokyo Hospital, Tokyo, Japan) [8] on a stand-alone personal computer running Windows (Microsoft, Redmond, WA, USA). ADC and FA were calculated pixel-by-pixel according to the conventional mono-exponential model from part of the q-space data, b-values of 0 and 1116 s/mm 2, because these data included multiple b-value data. Next, the full width at half maximum (FWHM) of the probability density function (PDF) was calculated as previously
Fig. 1. ROIs manually placed in the plaque (region 1), PWM (region 2), and NAWM (region 3) on a b = 0 diffusion-weighted image of a patient with MS.
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
627
Fig. 2. Another patient’s representative (A) b = 0 DTI (echo-planar T2-weighted image), (B) ADC, and (C) FA maps from conventional DTI data, and (D) RMSD map from QSI data. Corresponding T2- and T1-weighted images (E, F) at the same slice level for visual comparisons.
628
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629
Reproducibility was expressed in terms of the interrater correlation coefficient; the coefficient was 0.86 for the ADC analysis, 0.79 for the FA analysis, and 0.94 for the RMSD analysis. ADC values (mean ± 1 SD) for plaques, PWM, and NAWM were 0.640 ± 0.116, 0.545 ± 0.091, 0.490 ± 0.043 (10 − 3 mm 2 /s), respectively. FA values for plaques, PWM, and NAWM were 0.271 ± 0.072, 0.298 ± 0.080, 0.355 ± 0.092. RMSD values for plaques, PWM, and NAWM were 5.805 ± 1.201, 4.981 ± 0.857, 4.435 ± 0.400 μm, respectively. ADC values differed between plaques and PWM (P b 0.001) and between plaques and NAWM (P b 0.001). FA differed significantly (P b 0.001) between plaques and NAWM. RMSD data differed between plaques and PWM (P = 0.038), between plaques and NAWM (P b 0.001), and between PWM and NAWM (P = 0.019).
The third limitation of our study was the small number of patients evaluated and the lack of clinical correlations with diffusional metrics. FA and ADC values of the white matter can be influenced by duration and severity of MS. Therefore, before the usefulness of RMSD as an imaging biomarker can be established, longitudinal studies and correlations between RMSD and clinical disease characteristics must be established. In conclusion, RMSD values derived from QSI data may reflect microstructural changes and damage in the white matter of patients with MS with higher sensitivity than do ADC and FA values obtained from conventional DTI. More studies of the imaging–pathology relationship are needed, but QSI has the potential to provide new information for characterizing MS pathology in vivo. Conflict of interest statement
4. Discussion Our findings of highest ADC values and lowest FA values in plaques followed by PWM and NAWM are consistent with those of previous studies [1,24], and these patterns can be explained in part by the severity of white matter damage. In addition, RMSD values decreased from plaques to PWM and then NAWM; these changes varied significantly depending on the distance from the plaque. In a previous report addressing correlations between brain pathology and findings on imaging, the authors concluded that slight increases in ADC may be indicative of axonal loss, and decreases in FA may signal microglial activation in the white matter without plaques [25]. Our results showed that only RMSD was significantly different among plaques, PWM, and NAWM. Therefore, compared with conventional diffusion metrics, RMSD values from QSI may be a more sensitive biomarker to detect such graded pathologic change in white matter. The precise reason for the high sensitivity of RMSD in this regard remains unknown as yet. One explanation may lie in the fact that QSI uses multiple b-value data including high-b values (over 10000 s/mm2), which indicate intracellular water components, whereas conventional DTI is believed to measure water molecules in the extracellular space [6]. Moreover, QSI is a non-Gaussian diffusion analysis, with which it is possible (at least theoretically) to measure the full extent of water-molecule movement without having to assume Gaussian distribution of data, unlike the situation for conventional DTI. Therefore, QSI and its metric RMSD can lead to better estimation of actual neural tissue microstructural changes in vivo. One potential limitation of our study is the limited coverage obtained of the brain through QSI scanning (4 mm × 10 slices) and the relatively poor spatial resolution of 4-mm isovoxels. We used this condition to reduce the scan time to a clinically feasible duration. However, future investigations should focus on increasing both brain coverage and spatial resolution. Currently available techniques are limited in their ability to decrease scanning time on the MR scanners available in the clinical setting. However, various advanced techniques, such as compressed sensing [26], are expected to overcome this problem. Moreover, inherently lower SNR was expected in the calculated FA and ADC maps because they were calculated using data of only two b values and 6 motion probing gradient (MPG) axes and may substantially affect the results. Although we recognize the value of a greater number of MPG directions for more precise estimation of FA and ADC, we used the lower number because of the limited scan time for clinical practice. Another limitation is that QSI-derived P0 (probability for zero displacement) map was not used for the analysis in this study. We recognized that the P0 map was useful for MS lesion detection [6,19,27]. However, we thought that it was difficult to use P0 values for quantitative analysis because the values of P0 were usually scaled as arbitrary unit.
The authors declare that there are no conflicts of interest. Acknowledgments We thank Shuji Sato for help with data acquisition. This study was supported by a Grant-in-Aid for Scientific Research on Innovative Areas (Comprehensive Brain Science Network) from the Ministry of Education, Science, Sports, and Culture of Japan. This work was supported by JSPS KAKENHI Grant Number 24591788. References [1] Guo AC, MacFall JR, Provenzale JM. Multiple sclerosis: diffusion tensor MR imaging for evaluation of normal-appearing white matter. Radiology 2002;222:729–36. [2] Rovaris M, Bozzali M, Iannucci G, Ghezzi A, Caputo D, Montanari E, et al. Assessment of normal-appearing white and gray matter in patients with primary progressive multiple sclerosis: a diffusion-tensor magnetic resonance imaging study. Arch Neurol 2002;59:1406–12. [3] Rocca MA, Iannucci G, Rovaris M, Comi G, Filippi M. Occult tissue damage in patients with primary progressive multiple sclerosis is independent of T2-visible lesions–a diffusion tensor MR study. J Neurol 2003;250:456–60. [4] Inglese M, Bester M. Diffusion imaging in multiple sclerosis: research and clinical implications. NMR Biomed 2010;23:865–72. [5] Filippi M, Rocca MA, De Stefano N, Enzinger C, Fisher E, Horsfield MA, et al. Magnetic resonance techniques in multiple sclerosis: the present and the future. Arch Neurol 2011;68:1514–20. [6] Assaf Y, Ben-Bashat D, Chapman J, Peled S, Biton IE, Kafri M, et al. High b-value q-space analyzed diffusion-weighted MRI: application to multiple sclerosis. Magn Reson Med 2002;47:115–26. [7] Cohen Y, Assaf Y. High b-value q-space analyzed diffusion-weighted MRS and MRI in neuronal tissues – a technical review. NMR Biomed 2002;15:516–42. [8] Hori M, Fukunaga I, Masutani Y, Taoka T, Kamagata K, Suzuki Y, et al. Visualizing non-Gaussian diffusion: clinical application of q-space imaging and diffusional kurtosis imaging of the brain and spine. Magn Reson Med Sci 2012;11:221–33. [9] Jensen JH, Helpern JA, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005;53:1432–40. [10] Hori M, Motosugi U, Fatima Z, Kumagai H, Ikenaga S, Ishigame K, et al. A Comparison of Mean Displacement Values Using High b-Value Q-Space Diffusion-weighted MRI with Conventional Apparent Diffusion Coefficients in Patients with Stroke. Acad Radiol 2011;18:837–41. [11] Fatima Z, Motosugi U, Waqar AB, Hori M, Ishigame K, Oishi N, et al. Associations among q-space MRI, diffusion-weighted MRI and histopathological parameters in meningiomas. Eur Radiol 2013;23:2258–63. [12] Yamada K, Sakai K, Akazawa K, Sugimoto N, Nakagawa M, Mizuno T. Detection of early neuronal damage in CADASIL patients by q-space MR imaging. Neuroradiology 2013;55:283–90. [13] Kafri M, Sasson E, Assaf Y, Balash Y, Aiznstein O, Hausdorff JM, et al. High-level gait disorder: associations with specific white matter changes observed on advanced diffusion imaging. J Neuroimaging 2013;23:39–46. [14] Hori M, Motosug U, Fatima Z, Ishigame K, Araki T. Mean displacement map of spine and spinal cord disorders using high b-value q-space imaging-feasibility study. Acta Radiol 2011;52:1155–8. [15] Hori M, Fukunaga I, Masutani Y, Nakanishi A, Shimoji K, Kamagata K, et al. New diffusion metrics for spondylotic myelopathy at an early clinical stage. Eur Radiol 2012;22:1797–802. [16] Farrell JA, Smith SA, Gordon-Lipkin EM, Reich DS, Calabresi PA, van Zijl PC. High b-value q-space diffusion-weighted MRI of the human cervical spinal cord in vivo: feasibility and application to multiple sclerosis. Magn Reson Med 2008;59:1079–89.
M. Hori et al. / Magnetic Resonance Imaging 32 (2014) 625–629 [17] Farrell JA, Zhang J, Jones MV, Deboy CA, Hoffman PN, Landman BA, et al. q-space and conventional diffusion imaging of axon and myelin damage in the rat spinal cord after axotomy. Magn Reson Med 2010;63: 1323–35. [18] Raz E, Bester M, Sigmund EE, Tabesh A, Babb JS, Jaggi H, et al. A Better Characterization of Spinal Cord Damage in Multiple Sclerosis: A Diffusional Kurtosis Imaging Study. AJNR Am J Neuroradiol 2013;34:1846–52. [19] Assaf Y, Chapman J, Ben-Bashat D, Hendler T, Segev Y, Korczyn AD, et al. White matter changes in multiple sclerosis: correlation of q-space diffusion MRI and 1H MRS. Magn Reson Imaging 2005;23:703–10. [20] Yoshida M, Hori M, Yokoyama K, Fukunaga I, Suzuki M, Kamagata K, et al. Diffusional kurtosis imaging of normal-appearing white matter in multiple sclerosis: preliminary clinical experience. Jpn J Radiol 2013;31:50–5. [21] Fatima Z, Motosugi U, Hori M, Ishigame K, Kumagai H, Ikenaga S, et al. q-space imaging (QSI) of the brain: comparison of displacement parameters by QSI and DWI. Magn Reson Med Sci 2010;9:109–10.
629
[22] Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 1983;33:1444–52. [23] Polman CH, Reingold SC, Edan G, Filippi M, Hartung HP, Kappos L, et al. Diagnostic criteria for multiple sclerosis: 2005 revisions to the "McDonald Criteria". Ann Neurol 2005;58:840–6. [24] Assaf Y, Mayk A, Cohen Y. Displacement imaging of spinal cord using q-space diffusion-weighted MRI. Magn Reson Med 2000;44:713–22. [25] Moll NM, Rietsch AM, Thomas S, Ransohoff AJ, Lee JC, Fox R, et al. Multiple sclerosis normal-appearing white matter: pathology-imaging correlations. Ann Neurol 2011;70:764–73. [26] Menzel MI, Tan ET, Khare K, Sperl JI, King KF, Tao X, et al. Accelerated diffusion spectrum imaging in the human brain using compressed sensing. Magn Reson Med 2011;66:1226–33. [27] Pagani E, Bammer R, Horsfield MA, Rovaris M, Gass A, Ciccarelli O, et al. Diffusion MR imaging in multiple sclerosis: technical aspects and challenges. AJNR Am J Neuroradiol 2007;28:411–20.
