A Bayesian Spatio-Temporal Model for Functional Connectivity using Resting State fMRI and DTI data

Abstract

As non-invasive methods, functional magnetic resonance imaging (fMRI) plays an important role in studying the function of the human brain and diffusion tensor imaging (DTI) contributes to reveal the microstructure of the brain. Given the assumption that structural connectivity and functional connectivity are intercorrelated, we propose a spatio-temporal hierarchical Bayesian model for functional brain connectivity by jointly utilizing both resting state fMRI (rs-fMRI) and DTI data. In this model, structural information from DTI fused with ROI-level conventional functional connectivity serves as an informative prior for estimating composite functional connectivity. Moreover, indirect effect of structural connectivity is fused with direct effect of structural connectivity on composite functional connectivity, allowing us to take into account direct & indirect effect of structural information simultaneously. To assess the advantage of a joint DTI & rs-fMRI analysis, we compare our approach to the fMRI-only analysis in terms of the mean squared error (MSE) of estimators via simulated datasets. The results demonstrate that the DTI & rs-fMRI model reduces about 54.6% MSE of functional connectivity parameters compared to the conventional approach. We apply our model to analyze rs-fMRI and DTI data from 7 healthy subjects.