RECOVERY LIMITATIONS OF MEG SOURCE LOCALIZATION MODEL FOR EPILEPSY

Source localization in magnetoencephalography (MEG) has received a lot of attention in the recent years. MEG inverse problem is an ill-posed problem in from of B_ntimesl=G_ntimesmJ_mtimesl, where B is the vector of magnetic fields measured by the magnetic sensors, G is the lead field matrix, and J is the vector of source current in the brain. In practice, nLtm and J is usually a sparse vector. In this paper, we employ a complete reconstruction criterion to estimate the maximum number of active sources in the model or equivalently the volume of the active areas that guarantees a stable solution to the MEG inverse problem. Based on some prior information about the volume of the active areas, we then evaluate the stability of the solution with the employed model. Our approach guarantees an error upper bound on the solution with a certain number of active points. In the experiments, we have demonstrated the technique using some simulations.