A Bayesian Hierarchical Model for Detecting Group-Level fMRI Brain Activity

• Activity Number: 181
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xiao Yang*+
• Companies: University of Iowa
• Keywords: Bayesian ; hierarchical model ; GMRF ; fMRI

Abstract

Functional magnetic resonance imaging studies often involve analyzing data from multiple sessions and/or multiple subjects. The conventional Bayesian multilevel methods use a general linear model(GLM) at each level with different random effects variance components. Here, we propose a three-stage Bayesian hierarchical model to detect group-level activations of functional response using Gaussian Markov random fields (GMRF) priors on both activation location and magnitude for individual subjects. We use a binary variable to represent activation on each voxel. Doing this not only allows us to make inferences on subject level, but also help us to learn the patterns of response that are shared across groups. Inference is made based on Markov chain Monte Carlo (MCMC) methods. A simulation study is used to check the model applicability and sensitivity. Then, a real fMRI dataset shows different patterns of brain activity for two different groups.