Influence of Population Dependent Forward Models on Distributed EEG Source Reconstruction

In this study, we analyze how the forward model dependence on the study population influences the reconstruction of brain activity based on electroencephalographic (EEG) recordings. To this, we compare the source localization accuracy using generic and atlas-based head models, constructed with the Finite Difference Reciprocity method (FDRM). Additionally, we analyze the influence of including several tissues, as skull, scalp, gray matter, white matter, and cerebrospinal fluid. Comparison is carried out under a parametric empirical Bayesian (PEB) framework, that allows contrasting different forward modeling approaches using real data. Obtained results, based on event-related potentials (ERPs) of 31 subjects, show that the more realistic and more dependent on the study population the used head model, the better the ESI estimation.