The influence of realistic head geometry on EEG source localization

A fundamental problem in Electroencephalography (EEG) is the localization of sources in the brain responsible for generating the EEG signal measured on the scalp. This is referred to as the inverse problem. In order to solve for the location(s) of these source(s), one must model the source (cluster of neurons) and the volume conductor (head). Since the inverse problem has an infinite number of possible solutions, certain assumptions, such as the nature of the source and the geometrical representation of the volume conductor model, must be made. The degree to which sources can be localized accurately is dependent on such modeling assumptions, as well as EEG signal quality.; We simulate electric potentials on the surface of the scalp due to a current dipole in the brain of a realistically shaped head model (forward solution). Head-modeling errors are investigated by projecting these potentials onto a three-shell spherical head model and localization of the sources is determined (inverse solution). A similar technique is carried out using a standard “average” realistically shaped head model for inverse calculations. The results indicate that geometrical differences between the head model used in forward and inverse calculations are the dominant source of EEG source localization error, followed by signal noise and dipole depth.; In the final step, we localize the sources of the auditory evoked potential (AEP) using individual subject data (50 electrode sites) and grand-average data (5 subjects). The results show that localization errors due to grand-average EEG waveforms are slightly higher than errors caused by head-modeling mispecifications.; Our work suggests the use of individualized realistic head models with subject specific EEG data in order to accurately localize neural sources in the brain.