Dynamic Bayesian networks (DBNS) demonstrate impaired brain connectivity during performance of simultaneous movements in Parkinson's disease

摘要

Many symptoms of brain diseases may be caused by altered connectivity between brain regions, necessitating the development of suitable models for inferring effective connectivity in fMRI. Inspired by recent graphical approaches for inferring connectivity, here we propose dynamic Bayesian networks (DBNs) for learning the effective connectivity between a priori specified brain regions of interest (ROIs). We applied this method to fMRI data from Parkinson's disease (PD) and normal subjects performing a simultaneous movement task. Compared to the normal subject, the effective connectivity between motor regions was severely impaired in the PD subject, which was minimally ameliorated with L-dopa medication. These results imply a functional disconnection between brain regions far downstream from the basal ganglia, the initial site of pathology in PD. We suggest that DBNs provide a powerful framework to assess functional connectivity in fMRI studies of brain pathologies.