Explaining cognitive function in multiple sclerosis through networks of grey and white matter features: a joint independent component analysis.

Abstract

Cognitive impairment (CI) in multiple sclerosis (MS) is only partially explained by whole-brain volume measures, but independent component analysis (ICA) can extract regional patterns of damage in grey matter (GM) or white matter (WM) that have proven more closely associated with CI. Pathology in GM and WM occurs in parallel, and so patterns can span both. This study assessed whether joint-ICA of GM and WM features better explained cognitive function compared to single-tissue ICA. 89 people with MS underwent cognitive testing and magnetic resonance imaging. Structural T1 and diffusion-weighted images were used to measure GM volumes and WM connectomes (based on fractional anisotropy weighted by the number of streamlines). ICA was performed for each tissue type separately and as joint-ICA. For each tissue type and joint-ICA, 20 components were extracted. In stepwise linear regression models, joint-ICA components were significantly associated with all cognitive domains. Joint-ICA showed the highest variance explained for executive function (Adjusted R² = 0.35) and visual memory (Adjusted R² = 0.30), while WM-ICA explained the highest variance for working memory (Adjusted R² = 0.23). No significant differences were found between joint-ICA and single-tissue ICA in information processing speed or verbal memory. This is the first MS study to explore GM and WM features in a joint-ICA approach and shows that joint-ICA outperforms single-tissue analysis in some, but not all cognitive domains. This highlights that cognitive domains are differentially affected by tissue-specific features in MS and that processes spanning GM and WM should be considered when explaining cognition.