Temporal Dynamics of the Brain Using Variational Bayes Hidden Markov Models: Application in Autism

摘要

Investigating the functional connectivity (FC) patterns of the brain using resting-state functional magnetic resonance imaging (rs-fMRI) has been instrumental in revealing the effects of neurological disorders. Several studies have established that brain connectivity is dynamic in nature, and that brain diseases have an impact on both FC and its temporal properties. Various computational techniques have been proposed in the literature for modeling brain dynamics, yet most of these approaches have limitations that hinder the process of building accurate models. In this work, we explore a promising approach using Hidden Markov Models with Variational Bayesian Inference (VB-HMM) proposed by Ryali et al. (PLoS computational biology 12 (12), e1005138). A comprehensive study has been conducted quantifying useful statistical properties of the time-varying brain states and their underlying network configurations, providing insights on the influence of Autism on the functioning of the brain. This work focuses on the triple network model which consists of three major intrinsic connectivity networks (ICNs) that are known to play important roles in higher-order cognition. Autistic individuals demonstrated higher persistence in brain states possessing internetwork interactions in comparison to neurotypical subjects.