Year of Graduation
Classification of Normal and Pathological Brain Networks Based on Similarity of Graph Partitions
I consider a task of predicting normal and pathological phenotypes from macroscale human brain networks. These networks, called connectomes, represent connections between predefined brain regions. Thus brain regions becomes a network nodes, and aggregated neural pathways translates into edges of a such network. The set of brain regions is the same across different brains and the set of edges varies from subject to subject. I make use of the former property and assume that connectomes obtained from normal brains differs from those obtained from pathological in how brain regions cluster into communities. I demonstrate how dissimilarity in community structure could be used in a classification task on three real data sets.