Random Graph Mixture Model for Estimating the Shared Latent Structure in Groups of Brain Networks

• Activity Number: 375
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Christopher Bryant* and Hongtu Zhu and Joseph Ibrahim
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina
• Keywords: brain ; network ; Bayesian ; latent variable ; mixture model

Abstract

Networks offer a popular and powerful framework for assessing the relationships between large numbers of random variables. Many current scientific problems involve questions about not only the individual connections between nodes but also the overall structure of the network. These problems frequently arise in areas such as genomics (e.g. gene interaction networks), social sciences (e.g. social networks), and neuroimaging (e.g. functional brain networks). In this paper we present a novel Bayesian random graph mixture model for capturing the overlapping latent structure in brain networks estimated from functional magnetic resonance images (fMRI). We estimate the latent structure as well as group-specific graphical parameters via an efficient Gibbs sampling algorithm. We show via simulations that this method can accurately recover the clusters of individual networks as well as the graphical parameters governing the clusters. Ultimately we fit the model to real brain networks from the attention deficit hyperactivity disorder-200 (ADHD-200) sample.