Dynamic Functional Connectivity Change-Point Detection With Random Matrix Theory Inference

To study the dynamic nature of brain activity, functional magnetic resonance imaging (fMRI) data is useful including some temporal dependencies between the corresponding neural activity estimates. Recent studies have shown that the functional connectivity (FC) varies according to time and location which should be incorporated into the model. Modeling this dynamic FC (DFC) requires time-varying measures of spatial region of interest (ROI) sets. To know about the DFC, change-point detection in FC is of particular interest. In this paper, we propose a method of detecting a change-point based on the maximum of eigenvalues via random matrix theory (RMT). From covariance matrices for FC of all ROI's, the temporal change-point of FC is decided by an RMT approach. Simulation results show that our proposed method can detect meaningful FC change-points. We also illustrate the effectiveness of our FC detection approach by applying our method to epilepsy data where change-points detected are explained by the changes in memory capacity. Our study shows the possibility of RMT based approach in DFC change-point problem and in studying the complex dynamic pattern of functional brain interactions.