Joint density distributions of dynamic spatial brain networks show systematic variations at rest

Human resting-state functional magnetic resonance imaging data have been broadly studied previously to identify coherent spatio-temporal patterns of activity in functional brain networks and their dysfunction in brain disorders. While most studies focused on spatially static networks, here we developed an approach to estimate 4D spatially dynamic brain networks, evaluated systematic voxel-wise changes in such networks and the joint density distributions between pairs of networks using two-dimensional (2D) histograms. Clusters of 2D histograms computed using the k-means algorithm across subjects and sliding windows for each network pair showed significant group differences in subject-wise cluster occupancy and dwell time between healthy controls (CN) and patients with schizophrenia (SZ), implying altered network dynamics and interactions. This work provides unique insights into complex network-level relationships and possible dynamical mechanisms underlying SZ, and could potentially help in the development of novel diagnostics and biomarkers.