Inferring Dynamic Functional Connectivity from Field Potentials Using Graph Diffusion Autoregression

Abstract

Estimating dynamic functional connectivity (dFC) is attracting increased attention, spurred by rapid advancements in multi-site neural recording technologies and efforts to better understand cognitive processes. Yet, most studies focus on static estimates of functional connectivity that cannot capture highly dynamic neural processes, while also ignoring information about the structural organization of the brain. To address these issues, we introduce a class of network-constrained linear autoregressive models that give rise to a highly dynamic functional connectivity signal on the edges of a predefined structural connectivity graph. Furthermore, we demonstrate that adding an additional diffusion constraint improves the models performance. We successfully validated the resulting graph diffusion autoregressive (GDAR) model on simulated neural activity and recordings from subdural and intracortical micro-electrode arrays placed in macaque sensorimotor cortex demonstrating its ability to describe rapid communication dynamics induced by optogenetic stimulation, changes in resting state dFC following stroke and electrical stimulation, and neural correlates of behavior during a reach task.