Uncovering Latent Communication Patterns in Brain Networks via Adaptive Flow Routing

Unraveling how macroscopic cognitive phenotypes emerge from microscopic neuronal connectivity remains one of the core pursuits of neuroscience. To this end, researchers typically leverage multi-modal information from structural connectivity (SC) and functional connectivity (FC) to complete downstream tasks. Recent methodologies explore the intricate coupling mechanisms between SC and FC, attempting to fuse their representations at the regional level. However, while these approaches do incorporate useful neuroscientific observations, they predominantly operate at a topological or architectural level and lack a principled formulation grounded in neural communication dynamics. Consequently, they are limited in quantifying how information is actually routed between neural regions, and thus cannot fully explain why SC and FC exhibit dynamic states of both coupling and heterogeneity. In this paper, we formulate multi-modal fusion through the lens of neural communication dynamics and propose the Adaptive Flow Routing Network (AFR-Net), a physics-informed framework that models how structural constraints give rise to functional communication patterns, enabling interpretable discovery of critical neural pathways. Extensive experiments demonstrate that AFR-Net significantly outperforms state-of-the-art baselines. The code is available at \url{https://github.com/Skyyyy0920/AFR-Net}.

The human brain works as a network: billions of neurons are physically wired together by nerve fibers, and through this wiring different regions coordinate their activity to support thought and behavior. A central question in neuroscience — and a key to diagnosing brain disorders — is how the brain's fixed physical wiring gives rise to its flexible, ever-changing activity. Most AI methods that combine these two views treat the brain as a static shape and work as a black box, so they cannot explain how information actually travels between regions, or why their predictions are correct. We introduce AFR-Net, which treats the brain's wiring like an electrical circuit or a road network: every connection has a limited capacity, and regions that need to communicate send signals that spread along the most efficient routes. By estimating how much "traffic" each connection carries, the model predicts brain disorders more accurately than 14 existing methods — across anxiety, ADHD, obsessive–compulsive disorder, and Parkinson's disease — and, just as importantly, pinpoints the specific brain circuits that behave abnormally in patients, such as an over-active memory-to-self-reflection loop in anxiety. This turns a black-box predictor into a transparent tool that produces concrete, testable hypotheses for clinicians and neuroscientists.