Cortical EEG Source Localization of Focal Epilepsy

Abstract

Brain source localization allows us to localize different brain regions that are activated during neural activity. Several imaging modalities can be used for recording neural activity and are essential in clinical applications. One of these clinical applications is epilepsy diagnosis and localization. Structural or/and functional imaging techniques are used for patients to investigate epilepsy, classify seizures, and in pre-surgical evaluation. This report summarizes the most common imaging techniques for epilepsy diagnosis. It will then make use of electroencephalography (EEG) readings to localize epileptogenic regions in the brain, as it is a noninvasive technique with high temporal resolution. In addition, EEG requires low-cost hardware when compared with the other modalities such as functional Magnetic Resonance Imaging (fMRI), Positron Emission Tomography (PET), and Single Photon Emission Computed Tomography (SPECT). Moreover, the most common source models are discussed along with the used signal processing based techniques for source localization. In this work, distributed sources dipole model algorithms including the SAFFIRE and sLORETA are discussed and applied to simulated epileptic spikes. Upon examination of these algorithms, their potential in epilepsy source localization was proven with relatively low localization errors of 6.25 cm and 3.55 cm for sLORETA and SAFFIRE algorithms respectively. The SAFFIRE algorithm performance is investigated on epilepsy real data where the localized epileptogenic foci were consistent to the suggested locations by neurologists. Furthermore, the effect of reducing the number of electrodes on the source localization error was investigated on simulated epileptic spikes. The source localization error increased by 2.18 cm when reducing the number of electrodes from 256 down to 128. Then it increased by 3.7 mm when going from 128 electrodes to 64 electrodes. In conclusion, the localization error is inversely proportional to the number of electrodes used for recording brain potentials.