Evaluating Effective Connectivity in fMRI using Dynamic Granger Causality

Abstract

The invention of functional magnetic resonance imaging (fMRI) and blood oxygen level dependent contrast (BOLD) enable us to study how the brain works by non-invasive methods Granger causality analysis (GCA) is a popular method to analyze effective connectivity that has been widely used in neurosciences in the last decade We can use GCA to explore the interactions between the brain structures to find how the brain performs tasks and identify the hidden functional architecture However the application of GCA assumes that the analyzed time series must be covariance stationary (CS) it’s unlike the nature of nervous system that is dyanamic and time-varying We proposed a windowing-based Granger causality analysis to deal with upsampled non-CS time series called dynamic Granger causality analysis (DGCA) and verify the dynamic causal relationships in a simple auditory-motor task experiment The results show that the dynamic Granger causality analysis perform much more effective connectivity information than non-dynamic Granger causality analysis Our study demonstrate a new workflow to evaluate effective connectivity with upsampled data and windowing-based Granger causality analysis Accroding to the results of group analysis by clustering method we find out the patterns of the brain states while performing auditory-motor task

Date of Award	2014 Jan 23
Original language	English
Supervisor	Ming-Long Wu (Supervisor)