Springer Natural Computing
We provide a brief introduction to the nascent application of network theory to mesoscale networks in the human brain. Following an overview of the typical data-gathering, processing, and analysis methods employed in this field, we describe the process for inferring a graph from neural time series. A crucial step in the construction of a graph from time series is the thresholding of graph edges to ensure that the graphs represent physiological relationships rather than artifactual noise. We discuss the most popular currently employed methodologies and then introduce one of our own, based on the theory of random matrices. Finally, we provide a comparison of our random-matrix-theory thresholding approach with two dominant approaches on a data set of 1,000 real resting-state functional magnetic resonance imaging scans.