KESI-a novel method for spatial epileptic source localization in humans

BACKGROUND AND AIMS: Develop a novel method to assist presurgical evaluation in epileptic patients with pharmacologically intractable epileptic seizures, by spatial source localization of epileptic epicenters using stereoencephalography (SEEG) and electrocorticography (EcoG) recordings. METHODS: We developed kernel Electrical Source Imaging (kESI), which takes into account realistic brain morphology and spatial variations in brain conductivity. This method is parameter free, can localize multiple sources, and is flexible to allow arbitrary electrode positions. To account for the patient specific brain morphology, a patient’s MR scan can be used to evaluate the measured potential in a forward model using Finite Element Method in FEniCS software. The inhomogeneous electrical conductivity of the gray and white matter, skin and skull etc. can also be included. kESI is an inverse method, which relies on the construction of kernel functions requiring computation of the potentials generated in the brain by numerous basis functions covering the probed volume. This approach is based on our previous approach of kernel Current Source Density in 3 dimensions, while utilizing the patient specific forward modeling scheme above. RESULTS: To show the proof-of-concept we generated dipolar ground truth data in a simplified spherical brain model with uniform conductivity. We assumed the electrodes on the surface of the sphere and inside the spherical volume emulating ECoG and SEEG style recordings respectively. We show that the proposed method works, and can help in deciding how different distributions of electrodes affect the quality of reconstruction. CONCLUSIONS: kESI method facilitates accurate localization of the seizure onset zones, and a possible procedure for prescribing optimal distributions of electrodes depending on available prior knowledge (e.g. dysfunction of specific brain structures) and clinical resources (availability of specific electrodes, etc.).