SOS-MUSIC, a Subspace Approach for EEG Source Imaging Promoting Sparsity of Active Sources

Localizing multiple synchronous brain current sources from ElectroEncephaloGraphy (EEG) recordings is a challenging problem in presurgical evaluation of certain diseases such as drug resistant epilespy. In this paper, we propose a novel MUSIC-like (MUltiple SIgnal Classification) EEG source imaging method, named SOS-MUSIC. The latter minimizes the MUSIC metric as well as promoting Sparsity Of active Sources (SOS) to enhance performance. Indeed, by this way SOS-MUSIC helps to deal with synchronous (i.e. totally correlated) brain current sources, unlike classical approaches. This is illustrated through realistic computer simulations in the context of epilepsy by analyzing two challenging situations (low signal-to-noise ratios or correlated brain sources).