Beyond Thresholding: Fully-Weighted Graph Representations of Brain Functional Connectivity

Adam J. Schwarz¹, John McGonigle²

Functional connectivity analyses of fMRI data have leveraged recent advances in complex network theory, but these approaches have conventionally used a cut-off inter-node connection strength to threshold the network. This results in a sparse adjacency matrix amenable to conventional graph theoretic treatment, but requires the choice of a hard threshold (and verification of results over a range of such thresholds). We characterize the properties of fully-weighted human brain networks obtained by retaining all edges along with connection strength information, including the parametric dependence of a power law adjacency function (replacing the hard thresholding operation).