Attributed Graph Clustering for Functional Connectivity Networks

Network science has been successful at characterizing the topology of brain networks, showing alterations to the network structure due to disease and cognitive function. Functional connectivity networks (FCNs) represent different brain regions as the nodes and the dependency between them as the edges of a graph. One of the main tools to characterize the topology of FCNs is community detection. While different community detection methods have been applied to uncover the modular structure of the human brain, they rely only on the connectivity matrix. With the advances in graph signal processing, it is now possible to model the neuroimaging data as graph signals defined on the nodes of the FCN and the functional connectivity network as the underlying graph. In this paper, we present an optimal graph filter design procedure for identifying the community structure in FCNs. The resulting community detection algorithm employs both the connectivity and signal dynamics yielding a more robust community structure. The proposed method is applied to electroencephalogram (EEG) data collected from a study of error monitoring in the human brain.