A Novel Eigen-Volume-based Co-Activation Pattern Framework for Dynamic Functional Biomarkers of Multiple Sclerosis

Imaging biomarkers are essential for monitoring multiple sclerosis (MS), wand resting-state functional MRI (rs-fMRI) offers functional insights that complement structural imaging. This study investigates whether a novel co-activation pattern (CAP) approach for dynamic rs-fMRI can function as a dual-purpose biomarker in MS, aiding diagnosis and tracking disease severity. RS-fMRI scans from 25 relapsing-remitting MS patients and 41 healthy controls (HCs) were analyzed using a novel CAP-based approach. CAPs derived from individual time frames to capture dynamic brain activity patterns incorporated a bivariate similarity assessment, eigen volume-based dimensionality reduction, and consensus clustering. We evaluated the framework in two analyses: (1) a diagnostic evaluation, using dynamic CAP features-dwell time, persistence, and transition probabilities-for group comparisons and classification; and (2) a severity-prediction analysis, relating these CAP-derived measures to clinical disability (EDSS) in MS using LASSO regression. Method performance was benchmarked against standard CAP and sliding-window (SW) approaches. It revealed significant differences in brain activity between MS and HCs, within the default mode, sensorimotor, and language networks (p 0.75) and yielded better classification performance than standard CAP and SW approaches in classifying MS from HCs. These results suggest that dynamic brain activity patterns are altered in MS and linked to clinical disability. The proposed CAP provided improved performance in distinguishing MS patients, offering enhanced clinical monitoring. Transition probabilities emerged as a potential biomarker for tracking MS progression, with network shifts reflecting disease severity. As MS advances, increased transitions toward sensory, motor, and executive networks suggest compensatory recruitment. Conversely, reduced transitions from default mode and salience networks to sensorimotor and frontoparietal systems were associated with greater disability and diminished adaptive reorganization.