Solving the forward problem in EEG source analysis by spherical and fdm head modeling: a comparative analysis -...

Neural source localization techniques based on electroencephalography (EEG) use scalp potential data to infer the location of underlying neural activity. This procedure entails modeling the sources of EEG activity and modeling the head volume conduction process to link the modeled sources to the EEG, solving the so called EEG forward problem, and reconstructing the brain electrical activity from recorded EEG data, solving the EEG inverse problem. Many factors affect the accuracy of the forward and hence of the inverse problem solution, one of them is the shape of the head model. Realistic head models can lead to more accurate forward problem solutions, but imply heavier computational burdens in comparison to spherical models. Conversely, inverse solutions require the forward model to be computationally efficient. The aim of this study is to investigate the different general potentialities, in terms of EEG source reconstruction, which can be achieved adopting realistic or spherical geometries in head modeling. Previous studies in the literature analyzed the effect of head model geometry presenting results for particular cases of head models. In this paper, we re-address the effect of realistic geometry in head modeling, seeking for more general results by adopting the Montreal Neurological Institute (MNI) phantom model to represent a whole family of realistic head models. This paper presents results of a computer simulation study in which the potentialities of two different four-shell head models are compared, the realistic MNI-based FDM and the corresponding sensor-fitted spherical-shaped model, by means of the Point Spread Function (PSF) correlation maps, with a quantitative analysis of the accuracy in EEG source reconstruction given by head modeling refinement from the spherical to the more complex realistic FDM head modeling.