Abstract

Aberrant brain dynamics putatively characterize bipolar disorder (BD) and schizophrenia (SCZ). Previous studies often adopted a state discretization approach when investigating how individuals recruited recurring brain states. Since multiple brain states are likely engaged simultaneously at any given moment, focusing on the dominant state can obscure changes in less prominent but critical brain states in clinical populations. To address this limitation, we introduced a novel framework to simultaneously assess brain state engagement for multiple rain states, and we examined how brain state engagement differs in patients with BD or SCZ compared to healthy controls (HC). Using task-based data from the Human Connectome Project, we applied nonlinear manifold learning and K-means clustering to identify four recurring brain states. We then examined how the engagement and transition variability of these four states differed between patients with BD, SCZ, and HC across two other international, open-source datasets. Comparing these measures across groups revealed significantly altered state transition variability, but not engagement, across all four states in individuals with BD and SCZ during both resting-state and task-based fMRI. In our post hoc and exploratory analysis, we also observed associations between state transition variability and age as well as avolition. Our results suggest that disrupted state transition variability affects multiple brain states in BD and SCZ. By studying several brain states simultaneously, our framework more comprehensively reveals how brain dynamics differ across individuals and in psychiatric disorders.