Modeling versus Accuracy in EEG and MEG Data

this paper that highlight some of the problems of forward modeling, then discuss the impacts these results have on the inverse problem. We begin by assuming a perfect head model, that of the sphere, then show the lower bounds on localization accuracy of dipoles within this perfect forward model. For more realistic anatomy, the boundary element method (BEM) is a common numerical technique for solving the boundary integral equations. For a three-layer BEM, the computational requirements can be too intensive for many inverse techniques, so we examine a few simplifications. We quantify errors in generating this forward model by defining a regularized percentage error metric. We then apply this metric to a single layer boundary element solution, a multiple sphere approach, and the common single sphere model. We conclude with an MEG localization demonstration on a novel experimental human phantom, using both BEM and multiple spheres. Method