Estimate Information Flow in Brain Networks

• Activity Number: 623
• Type: Invited
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Graphics
• Author(s): Xi Luo* and Yi Zhao
• Companies: Brown University and Brown University
• Keywords: functional MRI ; multilevel models ; networks ; structural equation models

Abstract

The brain can be conceptualized as a network of connected nodes, and information, such as external stimuli, is processed while passing through the network. This talk studies the problem of inferring and visualizing information flow in brain networks from functional MRI data under event-related stimuli. We model such information flow as dynamic weights on the edges of the network. We develop a large-scale structural equation model to infer these weights, and we propose a constrained optimization approach for computation. We analyze the sufficient conditions to identify the model. We relax these conditions using integrated models for multilevel data. The numerical merits are illustrated using simulated data and a real fMRI dataset from a medium size cohort.