Comparison of performance of spherical and realistic head models in dipole localization from noisy EEG

The performance of a three-shell spherical head model versus the performance of a realistic head model is investigated when solving the inverse problem with a single dipole, in the presence of noise. This is evaluated by calculating the average dipole location error for 1000 noisy scalp potential sets, originating from the same test dipole and having the same noise level. The average location errors are obtained utilizing a local linearization, which is validated with a Monte-Carlo simulation. When the difference between the average location error utilizing a spherical and a realistic head model, represented by deltaR, is large for a large number of test dipoles, then it is worth using the more computationally demanding realistic head model. However, if deltaR is small for a large number of test dipoles, then it does not matter which model is used. For 27 electrodes, an electroencephalogram (EEG) epoch of one time sample and spatially white Gaussian noise, we found that the importance of the realistic head model over the spherical head model reduces by increasing the noise level. We further found that increasing the number of scalp electrodes from 27 to 44 has limited impact on the importance of the realistic head model over the spherical head model in EEG dipole source analysis. By increasing the number of time samples to six, the performance of the realistic head model in the inverse calculation gains importance compared with the three-shell spherical head model. Finally, we used spatially and temporally correlated background EEG instead of Gaussian noise. The advantage of the realistic head model over the spherical head model is reduced when applying correlated noise compared to Gaussian noise.