Multiple Sparse Priors applied to epileptic spike source localization

MA, CHENGXIN; LUU, PHAN; FENG, RUI; PAN, LI; FERNANDEZ CORAZZA, MARIANO; HU, JIE; TUCKER, DON

Conferencia; Live MEEG: Good Scientific Practices in EEG and MEG research; 2020

CNRS (France)

Introduction: About 30% of epileptic patients are non-responsive todrug treatment and neurosurgical resection of the epileptogenic tissue becomesthe primary therapeutic option. The goal in pre-surgical planning is todetermine the location of the so-called epileptogenic zone. This location canbe estimated by means of electroencephalography (EEG) source localizationalgorithms or ?inverse solvers?. Accurate source localization of epilepsyspikes is dependent on accurate electrical head models and appropriate inversesolvers. In this work, we compare the performance of the multiple sparse priors(MSP) inverse method with the standardized low-resolution brain electromagnetictomography (sLORETA) method in a set of fifteen epilepsy patients. Methods: For each subject, we built high-resolution andindividual electrical head models. The inverse solvers require the computationof the distribution of the electromagnetic field within the head. We solvedthis using our hexahedral finite element method (HexaFEM) engine. For eachpatient, spikes were grouped according to spatial similarity prior to spikeaveraging. The averaged spikes were source localized using our ownimplementations of the MSP and sLORETA methods. The criteria for accuracy wasgiven by the spatial mask of the volume resected in the surgery, consideringthat all subjects of this set remained seizure free. Results: The MSP method performed similarly to the sLORETAmethod and slightly better in terms of success rate. The MSP was more focalthan sLORETA with the advantage of not requiring an arbitrary selection of a hyperparameteror thresholding of reconstructed current density values to determine focus. Conclusions:Results suggest that both methods are complementary and could be used together.In practice, the MSP method will be easier to use and interpret compared tosLORETA.