Quantifying Dysregulation of fMRI-Derived Control Circuits for Computational Psychiatry

Abstract Psychiatric disorders are thought to result from dysregulated brain circuits, yet human neuroimaging currently lacks standardized methods for quantifying neural dysregulation. Here, we present a scalable framework for extracting fMRI-derived (generative) control circuits, then use circuit trajectories to estimate their control error. Using synthetic circuits, we first demonstrate that our framework accurately identifies each circuit's architecture and models its dynamics by estimation of transfer functions. As a use case, we then apply the framework to human task-based functional MRI data (UK Biobank, N=19,831). In a purely data-driven manner, without priors, our framework identified thalamus-linked prefrontal-limbic and ventral stream subcircuits, selectively engaged during sensorimotor processing of affective and non-affective stimuli. Finally, we demonstrate that circuit-wide dysregulation, defined by degree of drift from healthy trajectories, tracks symptom severity for neuroticism (ventral subcircuit), depression (prefrontal-limbic subcircuit), and bipolar disorder (full circuit).