Identifying Significant Regions of Brain Activation with the Use of fMRI Metadata

• Activity Number: 421
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Meredith Ray* and Hongmei Zhang and Jian Kang
• Companies: University of Memphis and University of Memphis and Emory University
• Keywords: Bayesian ; spatial analysis ; fMRI

Abstract

Coordinate-based functional magnetic resonance imaging (fMRI) meta data is gaining popularity due to its ability to identify regions in the brain that are activated by a given stimulant. Single fMRI studies contain a small number of subjects due to high costs; therefore, researchers turn to meta analysis to offset the issues of small samples such as low testing power and reproducibility. We elected to model the realization of meta data as a linear association with an overall mean while adjusting for some study effect. The mean was modeled as a mixture of unknown finite number of Dirichlet processes (DP) and adjusted for a study effect also modeled as a DP. Conditional on the mean and study effect, the random error follows a standard multivariate normal distribution. Modeling the mean as a mixture of DPs allows flexibility in capturing irregular spatial patterns and relaxes the typical normality assumptions. A Bayesian framework was implemented to draw model inferences. Simulation studies were conducted to explore the sensitivity and robustness of the method, but suggested a mediocre ability to correctly identify clusters. We also applied the proposed method to an emotion meta data.