GPU accelerated extraction of sparse Granger causality patterns

• Published in: Proceedings (International Symposium on Biomedical Imaging) pp. 604 - 607
• Main Authors: Sahoo, Dushyant, Honnorat, Nicolas, Davatzikos, Christos
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.04.2018
• Subjects: Brain modeling; Computational modeling; Extrapolation; Functional magnetic resonance imaging; GPU; Granger Causality; Graphics processing units; Parallel Computing; Proximal Alternating Linearized Minimization; Sociology; Statistics
• ISSN: 1945-8452
• DOI: 10.1109/ISBI.2018.8363648

Resting-state functional MRI, which provides a means to estimate the entire brain functional connectivity, has recently received a considerable amount of interest. This modality is increasingly being used to study functional connectivity dynamics, in particular with the aim of extracting individual biomarkers. However, the large amount of noise in the individual fMRI scans poses major challenges. In this work, we propose to analyze fMRI dynamics by extracting Granger causality patterns shared across subjects. This approach allows to capture individual brain organization while extracting population causality patterns which are more robust with respect to noise. We introduce an efficient method for the extraction of shared causality patterns, and we demonstrate its performance by processing the rs-fMRI scans of the hundred unrelated Human Connectome Project subjects.