Parallel Multilink Joint ICA for Multimodal Fusion of Gray Matter and Multiple Resting fMRI Networks.

In this study, we present a multimodal fusion approach, combining gray matter (GM) and multiple resting functional magnetic resonance imaging (fMRI) networks via a novel approach called parallel multilink joint independent component analysis (jICA) which combines 4D fMRI with 3D sMRI data. We focus on network-specific reconstruction and estimating joint relationship from differently distributed data by relaxing jICA assumption. Our methodology facilitates a detailed examination of altered connectivity patterns associated with Alzheimer's disease (AD). The study compares healthy controls (HC) and individuals with AD, employing two-sample t-tests with false discovery rate (FDR) correction to rigorously assess group differences. Network-specific correlation analysis reveals the joint relationships between different brain functions, allowing for a comprehensive exploration of AD pathology. Our approach also finds joint independent sources of altered activation patterns in key regions, such as the precuneus of the DMN, paracentral lobule of the sensorimotor domain, and cerebellum. This provides localized insights into the impact of AD on specific brain regions.