Structure-informed functional connectivity driven by identifiable and state-specific control regions

Abstract

AO_SCPLOWBSTRACTC_SCPLOWA challenge in neuroscience is to describe the contribution of the brain anatomical wiring to the emergence of coordinated neural activity underlying complex behavior. Indeed, patterns of remote coactivations that adjust with the ongoing task-demand do not systematically match direct, static anatomical links. Here, we propose that observed coactivation patterns, known as Functional Connectivity (FC), can be explained by a linear diffusion dynamics defined on the brain architecture and driven by control regions. Our model, termed structure-informed FC, is based on a novel interpretation of functional connectivity according to which different sets of brain regions controlling the information flow on a fixed anatomical wiring enable the emergence of state-specific FC. This observation leads us to introduce a framework for the identification of potential control centers in the brain. We find that well-defined, sparse and robust sets of control regions, which partially overlap across several task conditions and resting-state, produce FC patterns comparable to empirical ones. In conclusion, this work introduces a principled method for identifying brain regions underlying the task-specific control of brain activity. Significance statementUnderstanding how brain anatomy promotes particular patterns of coactivations among neural regions is a key challenge in neuroscience. This challenge can be addressed using network science and systems theory. Here, we propose that coactivations result from the diffusion of information through the network of anatomical links connecting brain regions, with certain regions controlling the dynamics. We translate this hypothesis into a model called structure-informed functional connectivity, and we introduce a framework for identifying control regions based on empirical data. We find that our model produces coactivation patterns comparable to empirical ones, and that distinct sets of control regions are associated with different functional states. These findings suggest that controllability is an important feature allowing the brain to reach different states.