Statistical Methods for finding Functional Connectivity

Abstract

We currently do not understand how the brain integrates information in real time to understand stimuli. The study of functional connectivity aims to understand which regions of the brain interact. However, current methods in functional connectivity are limited by requiring prior knowledge, such as specifying the number of clusters in the k-means algorithm, or confounding variables that result from simply thresholding the correlation matrix. We propose a method of sparse inverse covariance estimation paired with intersubject functional connectivity (ISFC) to overcome these challenges and give methods for selecting parameters. We show how our methods and prior algorithms perform on two datasets where we aim to understand how brain connections change through time and through changes in our internal motivations.