Recently in functional magnetic resonance imaging (fMRI) studies there has been an increased interest in understanding the dynamic manner in which brain regions communicate with one another, as subjects perform a set of experimental tasks or as their psychological state changes. Dynamic Connectivity Regression (DCR) is a data-driven technique used for detecting temporal change points in functional connectivity between brain regions where the number and location of the change points are unknown a priori. After finding the change points, DCR estimates a graph or set of relationships between the brain regions for data that falls between pairs of change points. In previous work, the method was predominantly validated using multi-subject data. In this paper, we concentrate on single-subject data and introduce a new DCR algorithm. The new algorithm increases accuracy for individual subject data with a small number of observations and reduces the number of false positives in the estimated undirected graphs. We also introduce a new Likelihood Ratio test for comparing sparse graphs across (or within) subjects; thus allowing us to determine whether data should be combined across subjects. We perform an extensive simulation analysis on vector autoregression (VAR) data as well as to an fMRI data set from a study (n = 23) of a state anxiety induction using a socially evaluative threat challenge. The focus on single-subject data allows us to study the variation between individuals and may provide us with a deeper knowledge of the workings of the brain.
Keywords: dynamic connectivity; functional connectivity; graph based change point detection; graphical lasso; network change points; stability selection; stationary bootstrap.