EEG Source Localization Based on Multiple fMRI Spatial Patterns

Abstract

EEG source localization is an ill-posed problem, and constraints are required to ensure the uniqueness of the solution. In this paper, using independent component analysis (ICA), multiple fMRI spatial patterns are employed as the covariance priors of the EEG source distribution. With the empirical Bayes (EB) framework, spatial patterns are automatically selected and EEG sources are estimated with Restricted Maximum Likelihood (ReML). The computer simulation suggests that, in contrast to the previous methods of EB in EEG source imaging, our approach is distinctly valuable in improvement of distributed source localization.