Functional Connectivity Analysis for Longitudinal fMRI

• Activity Number: 590
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Imaging
• Author(s): Mark Joseph Fiecas*+ and Ivor Cribben
• Companies: University of Warwick and Alberta School of Business
• Keywords: fMRI ; time series ; longitudinal data ; imaging

Abstract

Functional MRI (fMRI) studies could provide crucial information on the effects of rehabilitation on the neural mechanisms in patients with aphasia. Using fMRI, we are able to investigate cognitive and linguistic factors that contribute to the processing of written language, and how damage to specific brain regions affects these behavioural factors. In this study, using longitudinal fMRI acquired from subjects, we investigate how the functional network of the brain evolves as the subjects undergo various stages of computer-based therapies. Existing statistical methods have been lacking in the analysis of brain connectivity analysis using longitudinal fMRI data. We propose a linear mixed model using an l1 penalization for estimating a longitudinal network from fMRI data. The mixed model approach accounts for individual variability and within-individual covariability in the connectivity parameters, and the l1 penalty will account for the high-dimensionality of the data, which occurs due to the large number of brain regions considered in the analysis. We illustrate the utility of our method for estimating an evolving network structure across time induced by the different treatments.