EM Algorithms for Generalizing MCE and Focuss

The focal underdetermined system solver (FOCUSS) algorithm and the minimum current estimation (MCE) paradigm represent two ways of computing tomographic source reconstructions from MEG or EEG data. Both methods produce source estimates characterized by very compact, localized activity regions consistent with some neurophysiological evidence. While these methods have been studied extensively in the statistics and signal processing communities, they remain somewhat underutilized in the context of MEG and EEG source imaging. This paper presents simple extensions of both FOCUSS and MCE with the potential for handling flexible orientation constraints, estimation of distributed sources, and spatio-temporal smoothing. The resulting algorithms are easily implemented via the EM algorithm.