An Extended GEE Model with Latent Variables for Brain fMRI Connectivity

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Lijun Zhang*+ and Ming Wang and F. DuBois Bowman
• Companies: Emory University and Emory University and Emory University
• Keywords: fMRI ; Generalized Estimating Equation (GEE) ; brain connectivity

Abstract

Functional magnetic resonance imaging (fMRI) data provide us with a noninvasive tool to better understand cognitive processes associated with specific brain regions, and also reveal important functional properties that distinguish subgroups of subjects, e.g. a patient population from healthy control subjects. Subjects undergo scan sessions and are measured repeatedly over time, thus the fMRI data are collected longitudinally. To incorporate varied size of voxels within-region and the regional and temporal correlations across individuals, we propose a clustered weighted generalized estimating equation (CWGEE) with latent variables that produce consistent and robust regression estimates, leading to more accurate inferences regarding the functional correlations of brain connectivity. Also, we will compare our proposal with the two-stage general linear model which is one of the most common ways to analyze fMRI data to show the advantageous via simulation studies. Finally, our method will be taken into application from a real example of cocaine dependent study.