Functional and Structural Imaging Features of Cognitive Profiles in Multiple Sclerosis

Cognitive impairment affects 40%-70% of patients with multiple sclerosis (MS). Although recent hypothesis-driven and data-driven efforts have identified cognitive phenotypes, functional substrates underlying impairment in such phenotypes are still not clear. Our objective was to identify distinct cognitive phenotypes using clustering in a large cohort of patients with MS and characterize their associated clinical, structural MRI, and resting-state (RS) functional connectivity (FC) features.

We conducted an observational, cross-sectional study with right-handed patients with MS and age-matched and sex-matched healthy controls (HCs) who underwent neurologic and neuropsychological evaluation (Brief Repeatable Battery of Neuropsychological tests). Latent profile analysis on cognitive test Z-scores identified MS cognitive phenotypes. 3.0T MRI was used to derive atrophy measures and RS FC within default mode network (DMN I/II), executive control network (ECN), left/right fronto-parietal, language, and salience networks. Linear regression models compared structural MRI and RS FC across cognitive phenotypes and HCs, while polynomial contrasts assessed RS FC patterns across groups.

Three hundred and sixty-nine patients with MS and 168 HCs were enrolled. Four cognitive phenotypes were detected: preserved cognition ([PC], n = 67[18.2%]), mild single-domain verbal fluency involvement (MSD-VF, n = 37[10.0%]), mild single-domain attention involvement (MSD-A, n = 197[53.4%]), and severe multidomain (SMD) involvement (SMD, n = 68[18.4%]). At corrected threshold, PC patients showed decreased deep gray matter (GM) volumes (p < 0.001-0.001) and similar RS FC vs HCs, whereas MSD-VF exhibited no significant atrophy and no differences in RS FC vs HCs. MSD-A displayed lower cortical/deep GM volumes vs HCs (p < 0.001-0.005), whereas SMD had worse clinical and atrophy features (p < 0.001-0.04), as well as decreased DMN II RS FC (p = 0.01-0.04) vs other groups. Polynomial contrast analyses revealed significant quadratic patterns with increased RS FC in MSD-VF vs HC/PC groups, followed by decrease in MSD-A and SMD, within DMN II (p = 0.005), ECN (p = 0.005), and language network (p < 0.001).

PC exhibited intact RS FC, whereas SMD presented widespread structural and DMN RS FC damage. These findings advance our understanding of cognitive heterogeneity in MS by revealing distinct structural and functional neuroimaging profiles across cognitive phenotypes, encouraging future longitudinal and multimodal investigations to improve disease characterization and prognostic stratification.