MODELS OF THE HUMAN BRAIN AND THE SURROUNDING MEDIA - THEIR INFLUENCE ON THE RELIABILITY OF SOURCE LOCALIZATION

This article is a review of the evolution of models of the head as used in dipole source localization. Models fall into two classes: those that can be expressed in simple analytic form, such as the homogeneous sphere or spherical three-shell models, or those that can only be solved by numerical methods, such as the finite element approach. The latter models always involve heavy procedural and computational burdens. The trend over the last decade has been to use these more advanced models to estimate the error that would be incurred if one of the simpler spherical models were used instead for dipole source localization. An estimate is presented of the magnitudes of the random and systematic errors of localization that may be expected when using these methods.