EEG reconstruction and skull conductivity estimation using a Bayesian model promoting structured sparsity

Abstract

M/EEG source localization is an open research issue. To solve it, it is important to have good knowledge of several physical parameters to build a reliable head operator. Amongst them, the value of the conductivity of the human skull has remained controversial. This report introduces a novel hierarchical Bayesian framework to estimate the skull conductivity jointly with the brain activity from the M/EEG measurements to improve the reconstruction quality. A partially collapsed Gibbs sampler is used to draw samples asymptotically distributed according to the associated posterior. The generated samples are then used to estimate the brain activity and the model hyperparameters jointly in a completely unsupervised framework. We use synthetic and real data to illustrate the improvement of the reconstruction. The performance of our method is also compared with two optimization algorithms introduced by Vallagh\'e \textit{et al.} and Gutierrez \textit{et al.} respectively, showing that our method is able to provide results of similar or better quality while remaining applicable in a wider array of situations.