Grey Matter Atrophy in Multiple Sclerosis: Clinical Interpretation Depends on Choice of Analysis Method

Abstract

BackgroundStudies disagree on the location of grey matter (GM) atrophy in the multiple sclerosis (MS) brain.AimTo examine the consistency between FSL, FreeSurfer, SPM for GM atrophy measurement (for volumes, patient/control discrimination, and correlations with cognition).Materials and Methods127 MS patients and 50 controls were included and cortical and deep grey matter (DGM) volumetrics were performed. Consistency of volumes was assessed with Intraclass Correlation Coefficient/ICC. Consistency of patients/controls discrimination was assessed with Cohen’s d, t-tests, MANOVA and a penalized double-loop logistic classifier. Consistency of association with cognition was assessed with Pearson correlation coefficient and ANOVA. Voxel-based morphometry (SPM-VBM and FSL-VBM) and vertex-wise FreeSurfer were used for group-level comparisons.ResultsThe highest volumetry ICC were between SPM and FreeSurfer for cortical regions, and the lowest between SPM and FreeSurfer for DGM. The caudate nucleus and temporal lobes had high consistency between all software, while amygdala had lowest volumetric consistency. Consistency of patients/controls discrimination was largest in the DGM for all software, especially for thalamus and pallidum. The penalized double-loop logistic classifier most often selected the thalamus, pallidum and amygdala for all software. FSL yielded the largest number of significant correlations. DGM yielded stronger correlations with cognition than cortical volumes. Bilateral putamen and left insula volumes correlated with cognition using all methods.ConclusionGM volumes from FreeSurfer, FSL and SPM are different, especially for cortical regions. While group-level separation between MS and controls is comparable, correlations between regional GM volumes and clinical/cognitive variables in MS should be cautiously interpreted.

Whole‐brain atrophy assessed by proportional‐ versus registration‐based pipelines from 3TMRIin multiple sclerosis

“…Although lesion filling techniques improve accuracy of tissue segmentation (Chard, Jackson, Miller, & Wheeler‐Kingshott, 2010; González‐Villà et al., 2017), they did not appear to substantially affect measures of whole‐brain volume as we chose to use here (Dell'Oglio et al., 2015). There is an extensive literature regarding the optimization of GM versus WM tissue segmentation postprocessing techniques (Derakhshan et al., 2010; Popescu, Schoonheim, et al., 2016; Rocca et al., 2017), which is beyond the scope of this paper.…”

“…MRI‐derived volumetrics are prone to deviations throughout the data pipeline, including at the acquisition stage (e.g., head motion, hardware nonuniformity including magnetic field strength, gradient distortions, and pulse sequence type and parameters; Chu, Hurwitz, Tauhid, & Bakshi, 2017; Papinutto et al., 2017; Sharma et al., 2004; Shinohara et al., 2017) and segmentation procedure (e.g., preprocessing steps—inhomogeneity correction, method of tissue class segmentation, and normalization; Chard, Parker, Griffin, Thompson, & Miller, 2002; Chu, Hurwitz, et al., 2017; Durand‐Dubief et al., 2012; Granberg et al., 2016; Kazemi & Noorizadeh, 2014; Popescu, Schoonheim, et al., 2016; Vidal‐Jordana et al., 2017). Furthermore, brain volume may vary based on pathophysiological factors, including recent start of immunomodulatory therapy, acute inflammation, hydration status, time of day, tobacco use, genetics, and comorbid conditions (Rocca et al., 2017).…”

“…(2016) correlated postmortem, histopathologically defined cortical thickness with MRI‐acquired cortical gray matter (GM) measurements at 1.5T; the authors found statistically significant correlations only when using manually corrected (but not automated) pipelines in SIENAX and FreeSurfer. A separate study from the same group compared postprocessing pipelines in SIENAX, SPM, and FreeSurfer to evaluate the link between GM atrophy and cognitive performance in MS; although the software pipelines generally exhibited similar clinical correlations with cognitive variables, the authors found significant differences in deep GM and cortical structure measurements based, at least partly, on the choice of registration template/atlas (Popescu, Schoonheim, et al., 2016). The goal of this study was to compare the validity of two freely available widely used automated postprocessing algorithms for the assessment of normalized whole‐brain volume from 3T MRI.…”

Coordinate based random effect size meta-analysis of neuroimaging studies

“…The analysis of real data also provides an opportunity to qualitatively compare ClusterZ to the popular ALE and the ES-SDM algorithms. The meta-analysis of VBM studies of MS patients revealed some of the clusters that have been demonstrated to be the most consistently atrophied in a recent study comparing software packages for volumetric analysis in relapsing remitting MS [34]. A further example using real data analyses functional studies of healthy controls receiving painful mechanical stimulus.…”

Critical analysis on the present methods for brain volume measurements in multiple sclerosis

“…The same two platforms that gave these different results for teriflunomide (brain parenchymal fraction and SIENA) were considered to be the only reliable tools for assessing brain volume in a recent meta-analysis in which all the results from both of them were pooled 57 . If grey matter is the subject of investigation, variations are wide among methods used for analyses 35,36,37 , and several years of follow up may be necessary for conclusions 50 . On the other hand, cortical atrophy in MS occurs largely in a non-random manner and seems to affect distinct anatomical areas 58,59 .…”