Bayesian Joint Modeling of Multiple Brain Functional Networks

• Activity Number: 576 - Brain Connectivity Studies
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Joshua D. Lukemire* and Suprateek Kundu and Giuseppe Pagnoni and Ying Guo
• Companies: Emory University and Emory University and University of Modena and Reggio Emilia and Emory University
• Keywords: joint network estimation; brain networks; spike and slab prior; Stroop task

Abstract

Brain function is organized in networks exhibiting an intrinsic baseline structure with variations under different conditions. Existing approaches for uncovering this network structure typically do not explicitly model shared and differential patterns across networks, potentially reducing their detection power. We develop an integrative modeling approach for jointly modeling multiple brain networks. The proposed Bayesian Joint Network Learning (BJNL) approach develops flexible priors on the edge probabilities involving a common intrinsic baseline structure and differential effects specific to individual networks. Conditional on these edge probabilities, connection strengths are modeled under a Bayesian spike and slab prior on the off-diagonal elements of the inverse covariance matrix. The model is fit under a posterior computation scheme based on Markov chain Monte Carlo. Simulations illustrate that the proposed BJNL approach has increased power to detect differential edges and achieves greater accuracy in estimation of edge strengths compared to existing methods. An application to fMRI Stroop task data provides insights into brain network alterations between cognitive conditions.