Clustered-Temporal Bayesian Models for Brain Connectivity in Neuroimaging Data

Functional connectivity can be measured by considering co-activation of brain regions in resting-state functional magnetic resonance imaging. We explore contrasting fuctional connectivity between subjects with Autism Spectrum Disorder and controls using resting state fMRI data. Bayesian models are developed to explore differential connectivity using cross-correlated functional connectivity between region of interest pairs. Additionally, a regional-temporal model is proposed to directly model time sequence of resting state fMRI measurements at each brain region. We have implemented dynamic linear model to capture temporal structure of the data and the potential correlation between connected regions is modeled using hidden Potts model with latent variable. We apply the proposed models to analyse Autism Brain Image Data Exchange data set.