EEG-HR source localization is a non-invasive method to better delineate the source of the epileptic activity and thus the schema for electrodes implantation. Such source localization method relies on solving an inverse problem. In this study we evaluate the impact of bone conductivity, different head models and filtering, different methods of localization, the number of electrodes and the selection of different parts of the interictal epileptic spike.

Two groups of patients were studied. The first group (n=13) consists of patients with lesions on the MRI. They were recorded with 64 electrodes. The second group consists of non lesional patient but one (n=7) they were recorded with 128 electrodes. In the first group, the impact of bone conductivity, different head models, different filtering, and different methods of localization and the selection of different parts of the interictal spike was evaluate through the Euclidian distance of the resulting dipoles. In the second group, the impact of the number of electrodes (n=128/64/32) on the location of the dipole was evaluated.

The selection of the different parts of the interictal epileptic spikes and the different inverse methods (dipolar/scanning) did not return any significant difference on the source localization. Different filtering and bone conductivities did not significantly impact the localization. The different head model (realistic/realist) had significant impact on the localization of the source. Using 128 electrodes improved significantly the accuracy of the source localization.

Controlling the different parameters involved in the source localization significantly improve the accuracy of the inverse methods used to delineate the localization of the source of the IES and thus might be helpful to precise the schema of electrodes implantation.