Multimodal Brain Graph Analysis for Uncovering Path-Based Biomarkers in Schizophrenia: Integrating Static and Dynamic Approaches.

Analyzing multimodal brain networks offers new insights into the mechanisms of brain disorders. However, most studies focus on unimodal networks or use global/local graph metrics, often missing critical information about disrupted pathways. In this study, we propose a novel framework for estimating static and dynamic multimodal brain graphs by integrating time-resolved functional network connectivity (FNC) and gray matter features using a Gaussian graphical model. We then apply a path-based graph theory analysis to identify connectors and disconnectors, edges that respectively introduce or eliminate paths in the patient graph compared to controls. Our findings in schizophrenia reveal several edges linked to missing or additional paths, indicating network disruptions that are undetectable using unimodal or pairwise analyses. These disrupted pathways span both within- and cross-modal connections, particularly involving regions like the middle temporal gyrus and cerebellum. This work demonstrates the value of multimodal integration and emphasizes the importance of moving beyond individual edge-level analysis when characterizing disease-related network alterations. Our path-centric approach provides a complementary view of brain connectivity disruption and identifies candidate path-based markers of schizophrenia that warrant validation in independent datasets.