Bayesian MEG time courses with fMRI priors

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biophysics

Background:

• Magnetoencephalography (MEG) offers excellent temporal but limited spatial resolution for brain activity.
• Functional magnetic resonance imaging (fMRI) provides excellent spatial but limited temporal resolution.
• Integrating complementary neuroimaging modalities can yield more comprehensive insights into brain function.

Purpose of the Study:

• To implement a Bayesian framework for integrating fMRI data as spatial priors into MEG source reconstruction.
• To evaluate the effectiveness of fMRI spatial priors in improving MEG source localization accuracy.
• To assess the combined MEG and fMRI approach for enhanced neurobiological understanding, particularly in language processing.

Main Methods:

• Development of a Bayesian framework to incorporate fMRI spatial information into MEG inverse solutions.
• Validation using simulated MEG data encompassing both evoked and induced brain activity.
• Testing with experimental MEG data from sixteen human participants.

Main Results:

• Incorporating fMRI priors improved MEG spatial resolution by an average of 3 mm in simulated data.
• fMRI priors reduced spurious activation clusters in evoked activity and improved left-lateralization accuracy for induced activity in experimental data.
• Location error for induced sources in MEG data was significantly reduced by the use of fMRI spatial priors.

Conclusions:

• The use of fMRI spatial priors demonstrably enhances the accuracy and resolution of MEG source reconstruction.
• The combined MEG-fMRI approach offers superior spatial and temporal resolution for studying brain activity, including language neurobiology.
• Potential clinical applications include pre-surgical planning for epilepsy and evaluating language disorders in children.