A1161
Title: Fast Bayesian whole brain connectivity estimation by GPU-enhanced Gaussian processes
Authors: Yuting Mei - Vanderbilt University (United States)
Ilwoo Lyu - Ulsan National Institute of Science and Technology (Korea, South)
Kim Albert - Vanderbilt University Medical Center (United States)
Brian Boyd - Vanderbilt University Medical Center (United States)
Bennett Landman - Vanderbilt University (United States)
Warren Taylor - Vanderbilt University Medical Center (United States)
Hakmook Kang - Vanderbilt University (United States) [presenting]
Abstract: A previous Bayesian spatiotemporal model approach has been expanded to significantly reduce the computation burden by employing GPU (Graphics Processing Unit) computing and Gaussian Process to model the intra-voxel spatial correlation in each ROI (region of interest). A Bayesian double-fusion technique was used for enhancing the estimation of whole brain resting state functional connectivity (FC) based on functional magnetic resonance imaging (fMRI) data between brain regions by using structural connectivity (SC) based on diffusion tensor imaging (DTI) data. Concurrently acquired two imaging data are simultaneously used for FC estimation, which allows us to precisely investigate the relationship between FC and SC or alterations in white matter microstructural integrity. The method is applied to multi-subject data ($n = 45$) with depression ($n = 20$) and without depression ($n = 25$) to examine how FC differences are related to cognitive task performance in depression.
