Data-Driven Approach to Dynamic Resting State Functional Connectivity in Post-Traumatic Stress Disorder

Abstract

Posttraumatic stress disorder (PTSD) is a heterogenous psychological disorder that may result from exposure to a traumatic event. Using functional magnetic resonance imaging (fMRI), symptoms of PTSD have been associated with aberrations in brain networks that emerge in the absence of a given cognitive demand or task, called resting state networks. Most previous research in resting state networks and PTSD has focused on aberrations in the static functional connectivity among specific regions of interest (ROI) in the brain and within canonical networks constrained by a priori hypotheses. However, dynamic fMRI, an approach that examines changes in brain network characteristics over time, may provide a more sensitive measure to understand the network properties underlying dysfunction in PTSD. In addition, a data-driven analytic approach may reveal the contribution of other larger network disturbances beyond those revealed by hypothesis-driven examinations of ROIs or canonical networks. Therefore, the current study used a data-driven approach to characterize and subsequently compare brain network dynamics and recurrent connectivity states in a large sample of trauma exposed individuals (1,000+) with and without PTSD from the ENIGMA-PGC-PTSD workgroup. Static functional connectivity results showed those with PTSD had lower network efficiencies than Controls within and between sensorimotor and visual subnetworks. Further, network dynamics showed increased network efficiencies through the course of the scan for both groups, except in the visual subnetwork where those with PTSD showed blunted efficiencies through time. Those with PTSD also had fewer individual-level connectivity states, especially in the second half of the scan, compared to Controls suggesting a degree of stochasticity in the network over time. Finally, there were no group differences in dwell time or number of transitions of group-level connectivity states. Together, results suggest aberrancies in large-scale brain networks related to PTSD diagnosis beyond the most common analyzed ROIs. Unsurprisingly, in a large and heterogenous trauma sample, larger scale group results were not as robust compared to similar analyses in smaller homogenous trauma samples. Heterogeneity of PTSD, especially within diffuse brain networks, cannot be captured by evaluating only diagnostic groups, further work should be done to evaluate brain network dynamics with respect to specific symptoms and trauma types.