Bayesian State Space Modeling of Brain Effective Connectivity and Activation for fMRI Data

• Activity Number: 347
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Zhe Yu*+ and Hernando Ombao and Raquel Prado
• Companies: and University of California, Irvine and University of California, Santa Cruz
• Keywords: fMRI ; Bayesian ; State-space model ; HRF ; Effective connectivity

Abstract

To estimate effective connectivity and activation between brain regions of interest (ROI), we develop a state-space approach for fMRI data. The observation equation models the relation between fMRI signal and the BOLD response, and the state equation models connectivity through the time dependence structure of BOLD response amplitudes between ROIs. Existing methods ignore confounding effects from HRF, which could lead to erroneous conclusions in connectivity analysis. Our approach addresses the limitation by including estimation of region- and subject-specific HRF using linear basis functions. In addition, our approach is able to measure condition-specific connectivity by modeling the BOLD response explicitly for each condition. Furthermore, it captures group-level characteristics and inter-subject variability of connectivity by adding subject-level parameters in the model hierarchy. In this talk, we develop a Bayesian approach to inference and investigate the statistical properties of the approach through simulations. We also apply the method to data collected in a hand-motor task experiment from stroke patients, to explore special patterns of connectivity associated with stroke.