579439
research-article2015
MSJ0010.1177/1352458515579439Multiple Sclerosis JournalM Deppe, K Tabelow
MULTIPLE SCLEROSIS MSJ JOURNAL
Original Research Paper
Evidence for early, non-lesional cerebellar damage in patients with multiple sclerosis: DTI measures correlate with disability, atrophy, and disease duration
Multiple Sclerosis Journal 1–12 DOI: 10.1177/ 1352458515579439 © The Author(s), 2015. Reprints and permissions: http://www.sagepub.co.uk/ journalsPermissions.nav
Michael Deppe, Karsten Tabelow, Julia Krämer, Jan-Gerd Tenberge, Patrick Schiffler, Stefan Bittner, Wolfram Schwindt, Frauke Zipp, Heinz Wiendl and Sven G. Meuth
Abstract Background: Common symptoms of multiple sclerosis (MS) such as gait ataxia, poor coordination of the hands, and intention tremor are usually the result of dysfunctionality in the cerebellum. Magnetic resonance imaging (MRI) has frequently failed to detect cerebellar damage in the form of inflammatory lesions in patients presenting with symptoms of cerebellar dysfunction. Objective: To detect microstructural cerebellar tissue alterations in early MS patients with a “normal appearing” cerebellum using diffusion tensor imaging (DTI). Methods: A total of 68 patients with relapsing–remitting MS (RRMS) and without cerebellar lesions and 26 age-matched healthy controls were admitted to high-resolution MRI and DTI to assess microstructure and volume of the cerebellar white matter (CBWM). Results: We found cerebellar fractional anisotropy (FA) and CBWM volume reductions in the group of 68 patients. Interestingly, a subgroup of these patients that was derived by including only patients with early and mild MS (N=23, median age 30 years, median Expanded Disability Status Scale =1.5, median duration 28 months) showed already cerebellar FA but no CBWM volume reductions. FA reductions were correlated with disability, atrophy, and disease duration. Conclusion: “Normal appearing” cerebellar white matter can be damaged in a very early stage of RRMS. DTI seems to be a sensitive tool for detecting this hidden cerebellar damage.
Keywords: Diffusion tensor imaging, cerebellum, atrophy, degeneration, multiple sclerosis Date received: 22 July 2014; revised: 14 January 2015; 17 February 2015; accepted: 3 March 2015
Introduction Patients with multiple sclerosis (MS) may have a vast array of neurological symptoms where specific clinical signs may be related to the location of lesions within the central nervous system (CNS). Common symptoms include changes in sensation, muscle weakness, difficulties with coordination and balance (ataxia), or visual problems such as nystagmus and optic neuritis.1 Despite that, MS is traditionally considered as a white matter (WM) demyelinating disease, recent studies have underpinned extensive involvement of gray matter structures and the cerebellum already in early phases of the disease.2–4 The
cerebellum is known to be involved not only in motor but also in cognitive and affective processes. Cerebellar signs are significant contributors to the progression of disease disability and have a severe impact on life quality in affected patients. Derache et al.5 performed [18F]-fludeoxyglucose (FDG) positron emission tomography (PET) on patients with low disability and demonstrated a reduced metabolic rate of glucose (rCMRglu) predominantly in the thalamus and cerebellum in patients with relapsing–remitting multiple sclerosis (RRMS), which was significantly related to total lesion load. The authors suggested that hypometabolism might be caused by a reciprocal functional
Correspondence to: Michael Deppe Department of Neurology, Westfälische Wilhelms University, AlbertSchweitzer-Campus 1, Gebäude A1, 48149 Münster, Germany. [email protected] Michael Deppe Julia Krämer Jan-Gerd Tenberge Patrick Schiffler Stefan Bittner Heinz Wiendl Sven G. Meuth Department of Neurology, Westfälische Wilhelms University, AlbertSchweitzer-Campus 1, Gebäude A1, Münster, 48149, Germany Karsten Tabelow Weierstrass Institute, Berlin, Germany Wolfram Schwindt Department of Clinical Radiology, Westfälische Wilhelms University, Münster, Germany Frauke Zipp Department of Neurology, Rhine Main Neuroscience Network, Johannes Gutenberg University Medical Centre Mainz, Germany
http://msj.sagepub.com 1
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
Multiple Sclerosis Journal disruption between these structures. This interpretation is in line with several studies6 that have pointed out the importance of connecting loops between the cerebellum and frontal, superior temporal, limbic, and posterior parietal cortices in MS, and resting state functional magnetic resonance imaging (fMRI) studies reporting decreased connectivity principally between the cerebellum, thalamus, and cortex.7 Our study was motivated by the ability of DTI to detect functionally relevant (pathological) microstructural alterations on a quantitative basis by averaging fractional anisotropy (FA) values in regions of interest (ROI) of normal appearing WM and gray matter (GM) separately.3,8 However, assessing cerebellar diffusion alterations seemed to be more challenging when diffusion alterations have to be detected on a quantitative basis by averaging FA values in a ROI that outlines different tissue types.9 Cerebellar morphology comprises tightly folded cortex with underlying WM. Within the WM there are several deep nuclei embedded so that the whole cerebellum consists of spatially alternating different tissue types, each with different diffusion properties. These spatially alternating different tissue types are particularly problematic for DTI post-processing given that the necessary smoothing of the diffusion-weighted images10,11 typically blur and distort the structural and thus the diffusivity information in MR images. To account for this problem, Becker et al.12,13 developed a new algorithm for structural adaptive smoothing of diffusion weighted MRI data. The method, referred to as position-orientation adaptive smoothing (POAS), preserves edges of fine and anisotropic structures. In the present study, we employed POAS to investigate the extent and clinical correlations of early cerebellar involvement in patients with MS who were deemed to have no gross cerebellar lesions using conventional MRI sequences. In this context we aimed to test the following hypotheses: (1) Based on PET and fMRI findings, we hypothesized that cerebellar microstructural alterations can be demonstrated in patients with RRMS, even in patients without relevant obvious infratentorial lesions visible by conventional MRI. (2) Cerebellar alterations can be assessed on a quantitative basis by averaging the FA in a ROI outlining cerebellar WM. (3) Diffusion abnormalities of the cerebellum are related to factors associated with disease progression, such as WM lesion load as assessed by the total volume of WM black holes on T1-weighted (T1w) MRIs, disability as assessed by the EDSS, and disease duration. (4) Concerning hypothesis (1) our special interest was to find out which of the following diffusivity parameters would
be most sensitive to cerebellar white matter (CBWM) alterations in RRMS: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD). Methods Subjects Sixty-eight patients (45 f, 23 m) without evidence for cerebellar lesions were selected from 127 consecutive patients attending our clinic with relapsing–remitting MS (RRMS) (Expanded Disability Status Scale (EDSS): range 0–6.0, median 2.0, lower quartile 1.5, upper quartile 2.5; disease duration: range 4–313 months, median 70.5 months, lower quartile 35.5 months, upper quartile 135.5, age: median 36.0 years, range 19–56 years, SD 9.8 years) diagnosed according to revised McDonald criteria,14 were admitted to highresolution structural and diffusion weighted MRI at 3T. We derived from these patients a subgroup of patients with early and mild MS (inclusion criteria: duration < 50 months, EDSS ≤ 3.5; N=23, median/mean age 30.0/31.5 years, median EDSS 1.5, median duration 28 months). In the following we denote these 23 patients as the early RRMS group and the remaining 45 patients as the late RRMS group (median/mean age 37/39.1 years, median EDSS 2, median duration 113 months). Twentysix age-matched neurologically and psychiatrically healthy volunteers (19 f, 7 m) were included into the study as control participants (age: median/mean 30.5/37.0 years, range 23–69 years, SD 13.8 years, t-value: 0.18, p > 0.85); they were recruited by announcements in local newspapers. Written informed consent was obtained from all study participants. The participants were also informed that the examination could reveal potentially medically significant findings and given the option to request notification in the event of such findings. The interdisciplinary committee of the University of Münster and the Westphalia-Lippe Chamber of Physicians (Ärztekammer WestfalenLippe) approved all examinations (ref. 2010-378-b-S). MRI All participants were scanned using the same 3T Siemens TIM Trio MRI scanner and a 12-channel (matrix) head coil (Siemens AG, Erlangen, Germany) using the same MRI parameters and protocols. The following MRI sequences were obtained: a native isotropic (1.0 mm) 3D magnetization-prepared rapid gradient-echo imaging (MP-RAGE) T1w sequence and diffusion-weighted echo planar images for DTI (41 slices, slice thickness 3.6 mm, 20 gradient directions with b = 1000 s/mm2, 5 × b = 0 s/mm2).
2 http://msj.sagepub.com
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
M Deppe, K Tabelow et al.
Figure 1. Mean FA values of the cerebellar ROI for the healthy control subjects (CTRL), patients with early RRMS, and patients with late RRMS. Vertical bars denote 0.95% confidence intervals. The FA values are corrected for age with an age covariate mean of 36.6 years. The subpanel illustrates the two employed ROI for cerebellar and whole WM FA that did not overlap. The ROI are superimposed to the used FA template for spatial registration. The cerebellar ROI (brown) extends mainly across the superior, middle, and inferior peduncles, as well as cerebellar WM including cerebellar nuclei. The WM ROI covers mainly cerebral WM and the brain stem (green).
The following MRI sequences were applied only for the patients with MS to assess lesion load: an axial turbo spin-echo (TSE) Fluid-attenuated inversion recovery (FLAIR), a sagittal TSE FLAIR (32 slices, thickness 3.0 mm), and one 3D MP-RAGE T1w after intravenous gadolinium-DTPA (diethylene triamine penta-acetic acid) 0.1 mmol/kg injection.
DTI data post-processing All diffusion-weighted images were corrected for eddy currents and head movements using a recently developed algorithm15 and a multi-contrast image registration algorithm for the optimum spatial pre-processing of DTI data.16 The registration algorithm provided rigorous iterative multi-contrast registration steps based on FA contrasts and b0 contrasts (b = 0 s/mm2), so that volumetric effects in the resulting normalized FA maps were negligible.16 All registered diffusivity images corresponded to the MNI (Montreal Neurologic Institute) coordinate space. Smoothing of these images was performed by employing POAS12,13 using the ACID-toolbox (http://www.diffusiontools.org,)17 for SPM (http://www.fil.ion.ucl.ac.uk/spm/). POAS was parameterized by the following values: k* = 12, κ0 =
0.8, λ = 10, σ = 8, Ncoils = 1 (an explanation is given elsewhere12,17). For quantitative analysis of the DTI parameters we generated two regions of interest (ROI): one for cerebral WM and the other for CBWM (Figure 1, subpanel). The cerebral and cerebellar ROI were generated from spatially registered and averaged FA maps of 70 healthy control subjects by including either cerebral or cerebellar voxel with an average FA above 0.200 (typical for WM), respectively. To avoid any bias, the 70 subjects were not part of the control group used in the present study. The ROI were automatically created on the output images.18–20 For the quantitative DTI analysis we calculated the following diffusion parameters: If λ1 ⩾ λ2 ⩾ λ3 ⩾ 0 represent the sorted eigenvalues of the diffusion tensor, the following diffusion parameters were calculated from the individual tensor field maps: Fractional anisotropy FA =
1 2
(λ1 − λ2 ) 2 + (λ1 − λ3 ) 2 + (λ2 − λ3 ) 2 (λ12 + λ22 + λ32 )
,
mean diffusivity MD = (λ1 + λ2 + λ3 ) / 3, radial diffusivity RD= λ⊥ = (λ2 + λ3 ) / 2, axial diffusivity AD = λ|| = λ121. All DTI image-processing steps, i.e. registration,
http://msj.sagepub.com 3
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
Multiple Sclerosis Journal eddy current correction, smoothing, tensor estimation, diffusion parameter, and ROI calculation, were performed in a fully automated processing pipeline for quantitative assessment of DTI data (“Münster Neuroimaging Evaluation System” (EVAL)3,8,18). Used end points to assess disease progression A real ground truth measure that could be used as an objective reference for MS related microstructural alterations of the cerebellum on a histopathological level is not available for in vivo studies. Therefore we employed parameters (end points) that had been widely used in clinical trials to assess the severity of the patients’ MS: cerebral WM lesion load (WMLL), EDSS, WMVGMV ratio, cerebral WM FA, and disease duration. In our study WMLL, cerebral WM volumes (WMV), cerebral GM volumes (GMV), cerebellar WM volume (CBWMV), cerebellar GM volume (CBGMV), and intracranial volumes (ICV) were automatically estimated by means of FreeSurfer. The FLAIR and postgadolinium-DTPA images were diagnosed by an experienced neuroradiologist to exclude infratentorial lesions in the patients with RRMS. Structural MRI post-processing by FreeSurfer Prior to the automated analysis by FreeSurfer all 3D MPRAGE images were intensity inhomogeneity corrected to reduce segmentation errors using in-house software (Eval 3.0). CBGMV, CBWMV, ICV and WMLL were obtained from FreeSurfer (Version 5.1; http://surfer.nmr.mgh.harvard.edu/) in the same way as described previously.20 The amount of automatically detected hypointensities in the T1w images in units of mm3 served as a measure for WMLL. WM hyperintense lesions in T2w/FLAIR images were not considered because the main focus of the present study was the cerebellum. From the absolute volume of the cerebellum (unit: mm3) and the absolute ICV (unit: mm3) we calculated the relative (percentage) cerebellar WM volume as RCBWMV = CBWMV/ ICV × 100% and the relative cerebellar GM volume as RCBGMV = CBGMV/ICV × 100%. Statistical analysis To test our hypotheses, we employed univariate general linear models (GLM, Statistica 10, Stat Soft. Inc, www.statsoft.com) for cerebellar FA, RCBWMV, and RCBGMV with independent factors “group” (CTRL, early RRMS, late RRMS) and “age”, so that early vs. late RRMS group effects were not confounded by age. Mean FA values of the ROI, WM,
and GM volumes, and cerebral WM–GM ratios were compared between patients and controls by t-tests. We also employed GLM analyses to investigate correlations between EDSS, FA, and disease duration by accounting for the age dependency of FA and disease duration. Remark: We also performed exactly the same data analyses as described above but employed a conventional Gaussian smoothing kernel instead of POAS. Thereby it turned out that the new smoothing technique produced consistently higher statistical levels of inference. For the sake of clarity, we present in the following only the results derived by POAS. Results Early cerebellar FA reduction despite normal appearing infratentorial T1w/T2w MRI The GLM revealed significant cerebellar FA reductions in both, the early RRMS group and the late RRMS group relative to the control subjects (effect size: F(2, 90)=7.54, p
research-article2015
MSJ0010.1177/1352458515579439Multiple Sclerosis JournalM Deppe, K Tabelow
MULTIPLE SCLEROSIS MSJ JOURNAL
Original Research Paper
Evidence for early, non-lesional cerebellar damage in patients with multiple sclerosis: DTI measures correlate with disability, atrophy, and disease duration
Multiple Sclerosis Journal 1–12 DOI: 10.1177/ 1352458515579439 © The Author(s), 2015. Reprints and permissions: http://www.sagepub.co.uk/ journalsPermissions.nav
Michael Deppe, Karsten Tabelow, Julia Krämer, Jan-Gerd Tenberge, Patrick Schiffler, Stefan Bittner, Wolfram Schwindt, Frauke Zipp, Heinz Wiendl and Sven G. Meuth
Abstract Background: Common symptoms of multiple sclerosis (MS) such as gait ataxia, poor coordination of the hands, and intention tremor are usually the result of dysfunctionality in the cerebellum. Magnetic resonance imaging (MRI) has frequently failed to detect cerebellar damage in the form of inflammatory lesions in patients presenting with symptoms of cerebellar dysfunction. Objective: To detect microstructural cerebellar tissue alterations in early MS patients with a “normal appearing” cerebellum using diffusion tensor imaging (DTI). Methods: A total of 68 patients with relapsing–remitting MS (RRMS) and without cerebellar lesions and 26 age-matched healthy controls were admitted to high-resolution MRI and DTI to assess microstructure and volume of the cerebellar white matter (CBWM). Results: We found cerebellar fractional anisotropy (FA) and CBWM volume reductions in the group of 68 patients. Interestingly, a subgroup of these patients that was derived by including only patients with early and mild MS (N=23, median age 30 years, median Expanded Disability Status Scale =1.5, median duration 28 months) showed already cerebellar FA but no CBWM volume reductions. FA reductions were correlated with disability, atrophy, and disease duration. Conclusion: “Normal appearing” cerebellar white matter can be damaged in a very early stage of RRMS. DTI seems to be a sensitive tool for detecting this hidden cerebellar damage.
Keywords: Diffusion tensor imaging, cerebellum, atrophy, degeneration, multiple sclerosis Date received: 22 July 2014; revised: 14 January 2015; 17 February 2015; accepted: 3 March 2015
Introduction Patients with multiple sclerosis (MS) may have a vast array of neurological symptoms where specific clinical signs may be related to the location of lesions within the central nervous system (CNS). Common symptoms include changes in sensation, muscle weakness, difficulties with coordination and balance (ataxia), or visual problems such as nystagmus and optic neuritis.1 Despite that, MS is traditionally considered as a white matter (WM) demyelinating disease, recent studies have underpinned extensive involvement of gray matter structures and the cerebellum already in early phases of the disease.2–4 The
cerebellum is known to be involved not only in motor but also in cognitive and affective processes. Cerebellar signs are significant contributors to the progression of disease disability and have a severe impact on life quality in affected patients. Derache et al.5 performed [18F]-fludeoxyglucose (FDG) positron emission tomography (PET) on patients with low disability and demonstrated a reduced metabolic rate of glucose (rCMRglu) predominantly in the thalamus and cerebellum in patients with relapsing–remitting multiple sclerosis (RRMS), which was significantly related to total lesion load. The authors suggested that hypometabolism might be caused by a reciprocal functional
Correspondence to: Michael Deppe Department of Neurology, Westfälische Wilhelms University, AlbertSchweitzer-Campus 1, Gebäude A1, 48149 Münster, Germany. [email protected] Michael Deppe Julia Krämer Jan-Gerd Tenberge Patrick Schiffler Stefan Bittner Heinz Wiendl Sven G. Meuth Department of Neurology, Westfälische Wilhelms University, AlbertSchweitzer-Campus 1, Gebäude A1, Münster, 48149, Germany Karsten Tabelow Weierstrass Institute, Berlin, Germany Wolfram Schwindt Department of Clinical Radiology, Westfälische Wilhelms University, Münster, Germany Frauke Zipp Department of Neurology, Rhine Main Neuroscience Network, Johannes Gutenberg University Medical Centre Mainz, Germany
http://msj.sagepub.com 1
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
Multiple Sclerosis Journal disruption between these structures. This interpretation is in line with several studies6 that have pointed out the importance of connecting loops between the cerebellum and frontal, superior temporal, limbic, and posterior parietal cortices in MS, and resting state functional magnetic resonance imaging (fMRI) studies reporting decreased connectivity principally between the cerebellum, thalamus, and cortex.7 Our study was motivated by the ability of DTI to detect functionally relevant (pathological) microstructural alterations on a quantitative basis by averaging fractional anisotropy (FA) values in regions of interest (ROI) of normal appearing WM and gray matter (GM) separately.3,8 However, assessing cerebellar diffusion alterations seemed to be more challenging when diffusion alterations have to be detected on a quantitative basis by averaging FA values in a ROI that outlines different tissue types.9 Cerebellar morphology comprises tightly folded cortex with underlying WM. Within the WM there are several deep nuclei embedded so that the whole cerebellum consists of spatially alternating different tissue types, each with different diffusion properties. These spatially alternating different tissue types are particularly problematic for DTI post-processing given that the necessary smoothing of the diffusion-weighted images10,11 typically blur and distort the structural and thus the diffusivity information in MR images. To account for this problem, Becker et al.12,13 developed a new algorithm for structural adaptive smoothing of diffusion weighted MRI data. The method, referred to as position-orientation adaptive smoothing (POAS), preserves edges of fine and anisotropic structures. In the present study, we employed POAS to investigate the extent and clinical correlations of early cerebellar involvement in patients with MS who were deemed to have no gross cerebellar lesions using conventional MRI sequences. In this context we aimed to test the following hypotheses: (1) Based on PET and fMRI findings, we hypothesized that cerebellar microstructural alterations can be demonstrated in patients with RRMS, even in patients without relevant obvious infratentorial lesions visible by conventional MRI. (2) Cerebellar alterations can be assessed on a quantitative basis by averaging the FA in a ROI outlining cerebellar WM. (3) Diffusion abnormalities of the cerebellum are related to factors associated with disease progression, such as WM lesion load as assessed by the total volume of WM black holes on T1-weighted (T1w) MRIs, disability as assessed by the EDSS, and disease duration. (4) Concerning hypothesis (1) our special interest was to find out which of the following diffusivity parameters would
be most sensitive to cerebellar white matter (CBWM) alterations in RRMS: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD). Methods Subjects Sixty-eight patients (45 f, 23 m) without evidence for cerebellar lesions were selected from 127 consecutive patients attending our clinic with relapsing–remitting MS (RRMS) (Expanded Disability Status Scale (EDSS): range 0–6.0, median 2.0, lower quartile 1.5, upper quartile 2.5; disease duration: range 4–313 months, median 70.5 months, lower quartile 35.5 months, upper quartile 135.5, age: median 36.0 years, range 19–56 years, SD 9.8 years) diagnosed according to revised McDonald criteria,14 were admitted to highresolution structural and diffusion weighted MRI at 3T. We derived from these patients a subgroup of patients with early and mild MS (inclusion criteria: duration < 50 months, EDSS ≤ 3.5; N=23, median/mean age 30.0/31.5 years, median EDSS 1.5, median duration 28 months). In the following we denote these 23 patients as the early RRMS group and the remaining 45 patients as the late RRMS group (median/mean age 37/39.1 years, median EDSS 2, median duration 113 months). Twentysix age-matched neurologically and psychiatrically healthy volunteers (19 f, 7 m) were included into the study as control participants (age: median/mean 30.5/37.0 years, range 23–69 years, SD 13.8 years, t-value: 0.18, p > 0.85); they were recruited by announcements in local newspapers. Written informed consent was obtained from all study participants. The participants were also informed that the examination could reveal potentially medically significant findings and given the option to request notification in the event of such findings. The interdisciplinary committee of the University of Münster and the Westphalia-Lippe Chamber of Physicians (Ärztekammer WestfalenLippe) approved all examinations (ref. 2010-378-b-S). MRI All participants were scanned using the same 3T Siemens TIM Trio MRI scanner and a 12-channel (matrix) head coil (Siemens AG, Erlangen, Germany) using the same MRI parameters and protocols. The following MRI sequences were obtained: a native isotropic (1.0 mm) 3D magnetization-prepared rapid gradient-echo imaging (MP-RAGE) T1w sequence and diffusion-weighted echo planar images for DTI (41 slices, slice thickness 3.6 mm, 20 gradient directions with b = 1000 s/mm2, 5 × b = 0 s/mm2).
2 http://msj.sagepub.com
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
M Deppe, K Tabelow et al.
Figure 1. Mean FA values of the cerebellar ROI for the healthy control subjects (CTRL), patients with early RRMS, and patients with late RRMS. Vertical bars denote 0.95% confidence intervals. The FA values are corrected for age with an age covariate mean of 36.6 years. The subpanel illustrates the two employed ROI for cerebellar and whole WM FA that did not overlap. The ROI are superimposed to the used FA template for spatial registration. The cerebellar ROI (brown) extends mainly across the superior, middle, and inferior peduncles, as well as cerebellar WM including cerebellar nuclei. The WM ROI covers mainly cerebral WM and the brain stem (green).
The following MRI sequences were applied only for the patients with MS to assess lesion load: an axial turbo spin-echo (TSE) Fluid-attenuated inversion recovery (FLAIR), a sagittal TSE FLAIR (32 slices, thickness 3.0 mm), and one 3D MP-RAGE T1w after intravenous gadolinium-DTPA (diethylene triamine penta-acetic acid) 0.1 mmol/kg injection.
DTI data post-processing All diffusion-weighted images were corrected for eddy currents and head movements using a recently developed algorithm15 and a multi-contrast image registration algorithm for the optimum spatial pre-processing of DTI data.16 The registration algorithm provided rigorous iterative multi-contrast registration steps based on FA contrasts and b0 contrasts (b = 0 s/mm2), so that volumetric effects in the resulting normalized FA maps were negligible.16 All registered diffusivity images corresponded to the MNI (Montreal Neurologic Institute) coordinate space. Smoothing of these images was performed by employing POAS12,13 using the ACID-toolbox (http://www.diffusiontools.org,)17 for SPM (http://www.fil.ion.ucl.ac.uk/spm/). POAS was parameterized by the following values: k* = 12, κ0 =
0.8, λ = 10, σ = 8, Ncoils = 1 (an explanation is given elsewhere12,17). For quantitative analysis of the DTI parameters we generated two regions of interest (ROI): one for cerebral WM and the other for CBWM (Figure 1, subpanel). The cerebral and cerebellar ROI were generated from spatially registered and averaged FA maps of 70 healthy control subjects by including either cerebral or cerebellar voxel with an average FA above 0.200 (typical for WM), respectively. To avoid any bias, the 70 subjects were not part of the control group used in the present study. The ROI were automatically created on the output images.18–20 For the quantitative DTI analysis we calculated the following diffusion parameters: If λ1 ⩾ λ2 ⩾ λ3 ⩾ 0 represent the sorted eigenvalues of the diffusion tensor, the following diffusion parameters were calculated from the individual tensor field maps: Fractional anisotropy FA =
1 2
(λ1 − λ2 ) 2 + (λ1 − λ3 ) 2 + (λ2 − λ3 ) 2 (λ12 + λ22 + λ32 )
,
mean diffusivity MD = (λ1 + λ2 + λ3 ) / 3, radial diffusivity RD= λ⊥ = (λ2 + λ3 ) / 2, axial diffusivity AD = λ|| = λ121. All DTI image-processing steps, i.e. registration,
http://msj.sagepub.com 3
Downloaded from msj.sagepub.com at DALHOUSIE UNIV on May 17, 2015
Multiple Sclerosis Journal eddy current correction, smoothing, tensor estimation, diffusion parameter, and ROI calculation, were performed in a fully automated processing pipeline for quantitative assessment of DTI data (“Münster Neuroimaging Evaluation System” (EVAL)3,8,18). Used end points to assess disease progression A real ground truth measure that could be used as an objective reference for MS related microstructural alterations of the cerebellum on a histopathological level is not available for in vivo studies. Therefore we employed parameters (end points) that had been widely used in clinical trials to assess the severity of the patients’ MS: cerebral WM lesion load (WMLL), EDSS, WMVGMV ratio, cerebral WM FA, and disease duration. In our study WMLL, cerebral WM volumes (WMV), cerebral GM volumes (GMV), cerebellar WM volume (CBWMV), cerebellar GM volume (CBGMV), and intracranial volumes (ICV) were automatically estimated by means of FreeSurfer. The FLAIR and postgadolinium-DTPA images were diagnosed by an experienced neuroradiologist to exclude infratentorial lesions in the patients with RRMS. Structural MRI post-processing by FreeSurfer Prior to the automated analysis by FreeSurfer all 3D MPRAGE images were intensity inhomogeneity corrected to reduce segmentation errors using in-house software (Eval 3.0). CBGMV, CBWMV, ICV and WMLL were obtained from FreeSurfer (Version 5.1; http://surfer.nmr.mgh.harvard.edu/) in the same way as described previously.20 The amount of automatically detected hypointensities in the T1w images in units of mm3 served as a measure for WMLL. WM hyperintense lesions in T2w/FLAIR images were not considered because the main focus of the present study was the cerebellum. From the absolute volume of the cerebellum (unit: mm3) and the absolute ICV (unit: mm3) we calculated the relative (percentage) cerebellar WM volume as RCBWMV = CBWMV/ ICV × 100% and the relative cerebellar GM volume as RCBGMV = CBGMV/ICV × 100%. Statistical analysis To test our hypotheses, we employed univariate general linear models (GLM, Statistica 10, Stat Soft. Inc, www.statsoft.com) for cerebellar FA, RCBWMV, and RCBGMV with independent factors “group” (CTRL, early RRMS, late RRMS) and “age”, so that early vs. late RRMS group effects were not confounded by age. Mean FA values of the ROI, WM,
and GM volumes, and cerebral WM–GM ratios were compared between patients and controls by t-tests. We also employed GLM analyses to investigate correlations between EDSS, FA, and disease duration by accounting for the age dependency of FA and disease duration. Remark: We also performed exactly the same data analyses as described above but employed a conventional Gaussian smoothing kernel instead of POAS. Thereby it turned out that the new smoothing technique produced consistently higher statistical levels of inference. For the sake of clarity, we present in the following only the results derived by POAS. Results Early cerebellar FA reduction despite normal appearing infratentorial T1w/T2w MRI The GLM revealed significant cerebellar FA reductions in both, the early RRMS group and the late RRMS group relative to the control subjects (effect size: F(2, 90)=7.54, p
