Random location of multiple sparse priors for solving the MEG/EEG inverse problem

MEG/EEG brain imaging has become an important tool in neuroimaging. Current techniques based in Bayesian approaches require an a-priori definition of patch locations on the cortical manifold. Too many patches results in a complex optimisation problem, too few an under sampling of the solution space. In this work random locations of the possible active regions of the brain are proposed to iteratively arrive at a solution. We use Bayesian model averaging to combine different possible solutions. The proposed methodology was tested with synthetic MEG datasets reducing the localisation error of the approaches based on fixed locations. Real data from a visual attention study was used for validation.