Reproducible Coactivation Patterns of Functional Brain Networks Reveal the Aberrant Dynamic State Transition in Schizophrenia

Abstract

Abstract It is well documented that massive dynamic information is contained in the resting-state fMRI. Recent studies have identified recurring states dominated by similar coactivation patterns (CAP) and revealed their temporal dynamics. However, the reproducibility and generalizability of the CAP analysis is unclear. To address this question, the effects of methodological pipelines on CAP are comprehensively evaluated in this study, including preprocessing, network construction, cluster number and three independent cohorts. The CAP state dynamics are characterized by fraction of time, persistence, counts, and transition probability. Results demonstrate six reliable CAP states and their dynamic characteristics are also reproducible. The state transition probability is found to be positively associated with the spatial similarity. Furthermore, the aberrant CAP in schizophrenia has been investigated by using the reproducible method on three cohorts. Schizophrenia patients spend less time in CAP states that involve the fronto-parietal network, but more time in CAP states that involve the default mode and salience network. The aberrant dynamic characteristics of CAP are correlated with the symptom severity. These results reveal the reproducibility and generalizability of the CAP analysis, which can provide novel insights into the neuropathological mechanism associated with aberrant brain network dynamics of schizophrenia. Highlights Three coactivation patterns (CAPs) pairs with opposite coactivation profiles were identified, and the between-state transition probability was positively correlated with their spatial similarity. Good spatial and temporal reproducibility and generalizability of CAPs were achieved under varied analytic methods and independent cohorts. Schizophrenia patients showed altered temporal dynamics not only within the triple-network but also other primary and higher-order networks.