Rejuvenating Classical Source Localization Methods with Spatial Graph Filters.

EEG/MEG source imaging (ESI) aims to find the underlying brain sources to explain the observed EEG or MEG measurement. Multiple classical approaches have been proposed to solve the ESI problem based on different neurophysiological assumptions. To support the clinical decision making, it is important to estimate not only the exact location of source signal but also the boundary of extended source activation. Traditional methods usually render over-diffuse or sparse solution, which limits the source extent estimation accuracy. In this work, we exploit the graph structure defined in the 3D mesh of the brain by decomposing the spatial graph signal into low-, medium-, and high-frequency sub-spaces, and leverage the low frequency components of graph Fourier basis to approximate the extended region of source activation. We integrate the classical source localization methods with the low frequency subspace components derived from the spatial graph signal. The proposed method can effectively reconstruct focal extent patterns and significantly improve the performance compared to classical algorithms through both synthetic data and real EEG data.