Title: Time varying brain functional connectivity: Change-point estimation and testing
Authors: Jaehee Kim - Duksung Womens University (Korea, South)
DuBois Bowman - Columbia University (United States) [presenting]
Bin Cheng - Columbia University (United States)
Daniel Drake - Columbia University (United States)
Abstract: Functional connectivity (FC) methods are widely used to explore alterations in interregional associations that are linked to disease processes such as major depressive disorder (MDD) and Parkinson's disease (PD). There is emerging evidence toward time varying patterns of FC, although the vast majority of studies consider only static scalar-valued summaries of FC between pairs of brain regions. Many current approaches to dynamic FC are highly sensitive to a basic assumption regarding the choice of window length. We propose a CUSUM (or cumulative sum) analysis framework to assess time varying FC, with estimation and testing based on temporal profiles of scalar-valued functions of the sample covariance matrix. We examine the performance of our method using simulation studies and apply our method to determine change points in functional magnetic resonance imaging (fMRI) data from patients and healthy controls.