Spatial-temporal Bayesian inference for MEG/EEG

We recently described a new approach to the MEG/EEG inverse problem based on Bayesian Inference [1999]. Unlike almost all other approaches to the inverse problem, this approach does not result in a single "best" solution to the problem. Rather it yields a probability distribution of solutions upon which all subsequent inferences are based. This work demonstrated the utility of Bayesian inference both for including pertinent prior information (anatomical location and orientation, sparseness of regions of activity, limitations on current strength and spatial correlation) and for yielding robust results in spite of the under-determined inverse problem. This previous work focused on the analysis of data at a single point in time. We have extended the spatial-only analysis to a spatial-temporal Bayesian inference analysis of the full spatial-temporal MEG/EEG data set. Preliminary results of this extension are presented.