Online Program Home
My Program

Abstract Details

Activity Number: 657 - Statistical Network Models for Brain Connectivity Data Analysis
Type: Topic Contributed
Date/Time: Thursday, August 2, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Imaging
Abstract #329897
Title: Bayesian Integrative Analysis of Brain Functional Networks Incorporating Anatomical Knowledge
Author(s): Suprateek Kundu* and Ixavier Higgins and Ying Guo
Companies: Emory University Rollins School of Public Health and Rollins School of Public Health-Emory University and Emory University
Keywords: adaptive shrinkage; brain networks; Gaussian graphical models; multimodal imaging; Philadelphia Neurological Cohort; network reproducibility

In spite of the clear advantages, there exist some major challenges in integrative analyses of functional and structural brain imaging data, including the current incomplete understanding of structure-function relationship and limited accuracy in measuring anatomical structure. We propose a hierarchical Bayesian GGM approach which models the brain functional networks via sparse precision matrices having shrinkage parameters as random variables that are modeled using both anatomical structure as well as an independent baseline component. The proposed approach results in adaptive shrinkage and is flexible in identifying FC guided by structural connectivity knowledge, such that the functional network is appropriately informed by the anatomical knowledge but without being completely controlled by structural connections. This enables a robust brain network estimation even in the presence of mis-specified anatomical knowledge while accommodating heterogeneous relationships. We implement the approach via an efficient optimization algorithm which yields MAP estimates. Extensive numerical studies and an analysis of the PNC data reveal a clear advantage of the proposed approach.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program