Hierarchical region based ML EEG source reconstruction: A subspace projection approach

This paper presents a new method for EEG source reconstruction which is based on partitioning of cortical surface into a set of regions. The proposed method first takes advantage of subspace projection approach to determine most probable active regions in a hierarchical manner and then attempts to reach a current distribution confined to those regions. Simulation results with synthetic data show that the presented method achieves higher spatial resolution compared with previously proposed Weighted Minimum Norm (WMN) and Maximum Likelihood (ML) Approaches. The superiority of the new proposed method becomes more significant at low level of SNR especially when the sources are spread over several cortical regions.