Probabilistic Algorithm for Electromagnetic Brain Imaging with Spatio-Temporal and Forward Model Priors

In this paper we present a novel spatio-temporal inverse method for solving the inverse M/EEG problem. The contribution is two-folded; firstly, the proposed model allows for a sparse spatial and temporal source representation of the M/EEG by applying an automatic relevance determination type prior. The utility of a sparse spatio-temporal representation is based on the assumption that the underlying source activity is indeed sparse and smooth in time. Secondly, we seek to reduce the influence of forward model errors on the source estimates, by applying a stochastic forward model. Applying a stochastic forward model is motivated by the random noise contributions such as the geometry of the cortical surface and the electrode positions. Simulated data provide evidence that the spatio-temporal model leads to improved source estimates, especially at low signal-to-noise ratios, which is often the case in M/EEG.