Identifying and Characterizing Resting State Networks in Temporally Dynamic Functional Connectomes

An important application of resting state fMRI data has been to identify resting state networks (RSN). The conventional RSN studies attempted to discover consistent networks through functional connectivity analysis over the whole scan time, which implicitly assumes that RSNs are static. However, the brain undergoes dynamic functional state changes and the functional connectome patterns vary along with time, even in resting state. Hence, this study aims to characterize temporal brain dynamics in resting state. It utilizes the temporally dynamic functional connectome patterns to extract a set of resting state clusters and their corresponding RSNs based on the large-scale consistent, reproducible and predictable whole-brain reference system of dense individualized and common connectivity-based cortical landmarks (DICCCOL). Especially, an effective multi-view spectral clustering method was performed by treating each dynamic functional connectome pattern as one view, and this procedure was also applied on static multi-subject functional connectomes to obtain the static clusters for comparison. It turns out that some dynamic clusters exhibit high similarity with static clusters, suggesting the stability of those RSNs including the visual network and the default mode network. Moreover, two motor-related dynamic clusters show correspondence with one static cluster, which implies substantially more temporal variability of the motor resting network. Particularly, four dynamic clusters exhibited large differences in comparison with their corresponding static networks. Thus it is suggested that these four networks might play critically important roles in functional brain dynamics and interactions during resting state, offering novel insights into the brain function and its dynamics.