From Classical to Bayesian Estimators in the Interpretation of MEG and EEG

Abstract

The interpretation of MEG and EEG measurements often leads to solving a noisy, ill-conditioned matrix equation for the unknown source; the solution becomes an estimator of the true source. In practice, we need to know how to choose that estimator and what its properties are. In this paper, we present how a search for better estimators leads from the classical least-squares estimators through their regularized versions to Bayesian estimators. Though these estimators are rather well-known, the presented ’evolutionary’ path is less known and not easily found in the literature.