GWM-HFN: A Novel Gray-White Matter Heterogeneous Fusion Network for Functional Connectomes

Resting-state functional connectivity (FC) studies have predominantly centered on gray matter (GM), largely overlooking the functional contributions of white matter (WM). However, emerging evidence indicates WM blood-oxygen-level-dependent (BOLD) signals actively shape large-scale brain networks. Current methods to integrate WM face limitations, including challenges in assessing global network properties from bipartite GM-WM connections or the nascent stage of unified connectome models. Here, we introduce the Gray-White Matter Heterogeneous Fusion Network (GWM-HFN), a novel framework that defines GM-GM functional links via the covariance of their interaction profiles with WM bundles defined by a standardized atlas. Validated across six independent datasets, GWM-HFN demonstrates fair short-term (ICC ∼0.36) and slight-to-fair long-term (ICC ∼0.20) test-retest reliability, comparable to conventional GM-based FC. GWM-HFN exhibits distinct topological features, including small-worldness and enhanced modular segregation compared to GM-GM networks, capturing over 40% unique variance and an enhanced ability to resolve the unique connectivity patterns that differentiate individuals. Lifespan analyses reveal global linear declines and complex non-linear age effects in GWM-HFN connectivity, with peak connectivity in early adulthood (∼34 years). Clinically, individuals with autism spectrum disorder (ASD) show GWM-HFN-specific hyperconnectivity, which correlates with symptom severity and offers greater sensitivity than GM-GM FC. Furthermore, GWM-HFN connectivity patterns predict individual differences in cognitive performance, notably in language tasks. The GWM-HFN framework provides a robust and more comprehensive approach to understanding WM-mediated neural communication, integrating functional signals across both GM and WM, and offers promising avenues for developing neuroimaging biomarkers for aging and neuropsychiatric disorders.