Dynamic nodes, edges, and subnetworks in brain connectivity

Abstract

Recent brain research has been expedited by the network brain theory and neuroimaging methods for constructing whole-brain structural and functional brain networks. Functional brain networks are constructed based on the synchronous fMRI signal fluctuations among brain regions during resting state. Numerous studies have utilized functional brain network analysis to characterize individuals, to understand diseases and to test effects of treatments. All these applications of resting state functional brain networks were based on the assumption that functional brain networks are sufficiently stationary enough to describe a relatively long-lasting brain state. However, recent studies have shown dynamic natures of resting state brain networks within a relatively short time period. Thus, this study investigates dynamicity of network nodes, edges, and subnetworks of the whole brain using repeatedly measured fMRI data. We particularly focused on the hypothesis that integration among subregions of each node is highly dynamic, i.e., dynamic heterogeneity among voxels within a node. We also hypothesized that brain regions for highly dynamic membership for whole brain modules (high entropy) may correspond to hub regions. To test this hypothesis, we used resting state fMRI data from 12 healthy subjects measured at eight sessions during a 24-hour period. To evaluate the dynamicity of nodes, edges and other network properties, we used intra-class correlation (ICC). We found that highly stable node strength, node efficiency and clustering coefficient (ICC>0.5) at the bilateral superior parietal gyri, right precuneus, left hippocampus, and lateral inferior parietal lobule. We also found high entropy at bilateral parahippocampal gyri, bilateral hippocampus, and bilateral superior frontal gyri. These regions overlap with rich-club brain areas in previous studies. When we measured principal component analysis of each ROI time series, highly heterogeneous integration within ROI were found especially higher order brain regions. Furthermore, we examined the temporal consistency of effective connectivity of brain network submodules by looking at correlation between eight sessions. The higher order frontal area showed more dynamicity than lower sensory areas such as primary visual and auditory cortices. All these results reveal dynamic natures of the brain even during a 24-hour period. These dynamicity is not only node properties, edge strengths but also within node heterogeneous integrations, membership complexities and effective connectivities.