Dynamic Phase Synchrony based Ranked Spatio-Temporal Clustering for Tracking Time-Resolved Functional Brain Networks

Of late, there has been a growing interest in studying brain networks, particularly, for understanding spontaneous temporal changes in functional brain networks. Recently, phase synchrony based methods have been proposed to track instantaneous time-resolved functional connectivity without any need of windowing the data. This paper extends one such recently used phase synchrony measure in two steps. First, multiple temporal models are built from four-mode tensor that are further clustered to detect dynamic brain network communities. This clustering is based on spatio-temporal data and hence, is named as Spatio-Temporal Clustering (STC). Second, a method is proposed to rank all the communities allowing the proposed model to deal with multiple communities of differing time evolution. This helps in the comparison of network communities, especially, when available communities are too dense to provide relevant information for comparison. The ranking of communities allows for the dimensionality reduction of communities, while still maintaining the key brain networks. Intrinsic time-varying functional connectivity has been investigated for large scale brain networks, including default-mode network (DMN), visual network (VN), cognitive control network (CCN), auditory network (AN), etc. The proposed method provides a new complementary tool to investigate dynamic network states at a high temporal resolution and is tested on resting-state functional MRI data of 26 typically developing controls (TDC) and 35 autism spectrum disorder (ASD) subjects. Simulation results demonstrate that ASD subjects have altered dynamic brain networks compared to TDC.