Adapted K-Core Decomposition and Visualization for Functional Magnetic Resonance Imaging Connectivity Networks

Medical imaging modalities, such as functional magnetic resonance imaging (fMRI) are being increasingly used to study the human brain. Analysis of the images has led to findings describing diseases, such as schizophrenia and post-traumatic stress disorder. One of the most widely used methods of analysis involves creating functional connectivity network (FCN) abstractions. These summarize the temporal relationships between regions of interest (ROIs) in the brain and can be used to easily compare subjects, e.g. healthy against schizophrenia. Visual analytics is widely used to facilitate such analysis, with existing approaches designed to enable and simplify detailed interpretation of single networks and pairs of networks in comparison. Prior to such detailed analysis, grouping and aggregation is often performed on the data, which is a time consuming and difficult task. Existing methods for doing this are commonly statistical, while others visualize the cohort without presenting vital network details of the individual FCNs. Thus, there is an opportunity for alternative visual analytics to facilitate the grouping by incorporating the network details. Graph decomposition, such as k-core decomposition, can be used to simplify the representation of networks, while retaining these vital network details. In this study, we propose an adapted k-core decomposition algorithm and visualization, which calculates the connected component information of nodes in the FCNs, a key detail in analysis. Our visualization combines this information with the decomposition to display more details about FCNs at a high-level than contemporary approaches. We present a prototype of our method, demonstrating the ability to group and aggregate the data without the loss of vital network details for further detailed analysis.