An in-vivo comparison of source localization methods

Electrical source imaging (ESI) aims at reconstructing the electrical brain activity from measurements of the electric field on the scalp. ESI is a key element in the analysis of EEG data, in both research and clinical settings. In the last twenty years several algorithms have been applied for solving the ill- posed EEG inverse problem. Most of these popular methods can be derived within a Bayesian statistical framework in which all variables can be modelled as random variables with associated probability density functions (pdf) and the solution of the inverse problem is the posterior pdf for the unknown primary current distribution conditioned on the measurements. The different methods mainly differ from each other by the quality and quantity of a priori information they use in order to solve the EEG inverse problem. In this study [1] we validate and compare ten different ESI methods (wMNE, dSPM, sLORETA, eLORETA, LCMV, dipole fitting, RAP-MUSIC, MxNE, gamma map and Sesame) "in vivo", by exploiting a recently published EEG dataset [2] for which the ground truth is known. We compare the different inverse methods under multiple choices of input parameters, to assess the accuracy of the best reconstruction, as well as the impact of the parameters on the localization performance