The influence of skull parameters in a realastically shaped head model on the accuracy of EEG dipole localization

Magnetic resonance imaging (MRI) provides the means for the generation of head models with a high geometrical detail. Errors in the generation of realistically shaped models are likely to be made, due to the identification of the different anatomical structures. The poorly conducting skull layer plays a special role, since it is known to have a large effect on the scalp potentials and is difficult to distinguish in MRI. If source localisation is applied to EEG, then the systematic errors in the parameters of the reconstructed sources are partly due to the misspecifications of the head model. - In this paper, the influence of certain head model parameters on the systematic errors of reconstructed dipole sources is investigated. Variations in the skull conductivity and thickness, as well as local variations in the skull thickness, are considered. In order to do so, the sensitivity measure is introduced, which specifies the amount of change of a dipole parameter due to a specific model variation. Sensitivity maps are constructed for layers of dipoles underneath the brain surface. The maps of dipole sensitivities show the local distribution of the systematic errors to be expected. The computations are performed by means of a derivative method, which utilizes a linearization of the forward problem.