Diffusion tensor imaging (DTI) provides connectivity information that helps illuminate the processes underlying normal development as well as brain disorders such as autism and schizophrenia. Researchers have widely adopted graph representations to model DTI connectivity among brain structures; however, most measures of connectivity have been centered on nodes, rather than edges, in these graphs. We present an edge-based algorithm for assessing anatomic connectivity; this approach provides information about connections among brain structures, rather than information about structures themselves. This perspective allows us to formulate multivariate graph-based models of altered connectivity that distinguish among experimental groups. We demonstrate the utility of this approach by analyzing data from an ongoing study of schizophrenia.
Keywords: Data mining; Diffusion tensor imaging; Network analysis; Schizophrenia; Tractography.