Dipole localization accuracy using grand-average EEG data sets

Objective: Dipole localization of grand-average event related potentials only give a tentative description of the estimated underlying neural sources. This study evaluates the differences in dipole solutions between individual and group-average data sets using a standard realistic head model. Methods: Auditory evoked potentials were recorded from 14 right-handed healthy subjects using a 64 electrode montage. Inverse dipole solutions were obtained for each individual data set, as well as for all individual responses averaged together (grand-average). Differences in dipole solutions between individual and grand-average responses are reported. Simulations using a two dipole model with 15 different electrode sets are then used to investigate the effects of electrode misplacement and random noise on dipole localization. These effects are compared to those due to grand-averaging. Results: The average differences in dipole locations between the individual and grand-averaged data sets were approximately 1.1 cm (SD=0.7 cm). This difference is larger than typical localization errors due to electrode misplacement and typical noise. Conclusions: Using a standard realistic head model, it is concluded that dipole solutions based on group-averaged EEG datasets are significantly different than those obtained using subject-specific data.