Bayesian MEG time courses with fMRI priors

Abstract

AbstractMagnetoencephalography (MEG) records brain activity with excellent temporal and good spatial resolution, while functional magnetic resonance imaging (fMRI) offers good temporal and excellent spatial resolution. The aim of this study is to implement a Bayesian framework to use fMRI data as spatial priors for MEG inverse solutions. We used simulated MEG data with both evoked and induced activity and experimental MEG data from sixteen participants to examine the effectiveness of using fMRI spatial priors in MEG source reconstruction. For simulated MEG data, incorporating the prior information from fMRI increased the spatial resolution of MEG source reconstruction by 3 mm on average. For experimental MEG data, fMRI spatial information reduced the spurious clusters for evoked activity and showed more left-lateralized activation pattern for induced activity. The use of fMRI spatial priors greatly reduced location error for induced source in MEG data. Our results provide empirical evidence that the use of fMRI spatial priors improves the accuracy of MEG source reconstruction. The combined MEG and fMRI approach can provide neuroimaging data with better spatial and temporal resolutions to add another perspective to our understanding of the neurobiology of language. The potential clinical applications include pre-surgical evaluation of language function for epilepsy patients and evaluation of language network for children with language disorders.