Abstracts

Rationale: EEG source localization is an efficient tool for presurgical planning in case of medically intractable partial epilepsy. It is inexpensive and available in most clinical settings. One major concern about EEG recording is its low spatial resolution. This can be deteriorated by insufficient recording channels, i.e. low resolution EEG recording systems with 32 electrodes or less. The aim of this study is to look at the role that the number of recording channels can play in determining the accuracy of the EEG source localization by comparing the inverse solution to clinical findings such as surgical resection and seizure onset zone (SOZ) from intracranial recordings. Methods: Computer simulation and patient data was analyzed to study the relation between localization error and number of electrodes. Five pediatric patients suffering from temporal and extra-temporal partial epilepsy were included in this study. The interictal spikes from the pre-surgical EEG recording of these patients were extracted and a realistic geometry boundary element method (BEM) model was built for each individual patient based on the pre-operative MRI. The inverse solution was calculated for different electrode montages, i.e. 32, 64, 96, 128 electrodes, and the results were compared with the surgical resection which was extracted from post-operative MRI and the SOZ electrodes, extracted from CT images and based on physician's reports. The original recording was recorded with 128 electrodes. To obtain montages with lower density, electrodes were removed in a manner to guarantee the uniformity of the electrodes over scalp. Results: The localization error is defined in two manners. First, the distance between the location of the inverse-solution maximum to resection border and second, by projecting the inverse solution to the cortex and then calculating the distance between the location of the maximum with the set of SOZ electrodes. The localization error decreases when the number of electrodes increases but the improvement in localization decreases with increasing electrode number. This plateauing effect suggests that increasing the electrode numbers indefinitely will not improve the localization much. A set of computer simulations were performed to further study this plateauing effect. Random locations within a realistic head model were selected and dipoles with random orientation were simulated. Various noise power levels were added to the simulation and the inverse was solved for the aforementioned electrode configurations and the localization error was calculated. Same effects can be seen. Conclusions: A systematic study using patient data and computer simulations was performed to investigate the relation between localization accuracy and number of recording channels in EEG. The results suggest that using high resolution EEG recording systems, i.e. 64 channels or more, is necessary for reliable source localization. It was also shown that the localization accuracy improves when electrode numbers are increased, but the improvement rate ultimately slows down (This work was supported in part by NIH EB006433).