Capturing Spatial Dynamics Using Time-Resolved Referenced-Informed Network Estimation Techniques.

We recently showed that the brain is spatially dynamic, i.e., the spatial patterns of brain networks evolve over time. Yet, most studies, even those studying brain dynamics, omit this information, leading to incorrect inferences. Moreover, spatial dynamics carry unique information about networks hidden from existing spatially static approaches. However, estimating networks in a time-resolved manner is challenging because (1) resting-state fMRI (rsfMRI) has a low signal-to-noise ratio (SNR), (2) relatively short time segments have insufficient information to characterize the spatiotemporal patterns stand-alone, and (3) identification of correspondence within and between-subject is suboptimal and uncertain. Inspired by the group-inference framework, we addressed these limitations by adopting a referenced-informed estimation technique in a time-resolved manner to capture time-varying brain networks and their spatial functional network connectivity (spFNC). Our approach detects schizophrenia alterations in dynamic spFNC. Interestingly these changes in low-dimensional global brain dynamics are also manifested in high-dimensional (voxel-level) space.