부산대학교 도서관

Epilepsy surgery still lacks an operational method for finding the epileptogenic zone (EZ) or the minimal amount of tissue that has to be resected in order to leave the patient seizure-free. Here we propose a method for predicting the result of a resection from data collected before the eventual resection, therefore allowing an optimal surgical planning. Our major hypothesis is that focal generalized epilepsies are caused by sub-systems with excessive afferent connectivity to and from the rest of the neuronal tissue. To address the issue of delineating the EZ before the actual resection, we propose a paradigm performing "virtual surgery" on the matrix of connectivity between local EEG measurements. The virtual resection removes not only the nodes covered by the suspected EZ and their connections but also subtracts the influence of these nodes on the rest of connectivity. The residual connectivity is then compared to the original one and a significant decrease indicate that the resection contains the EZ or at least large part of it. We tested this approach on a computational model of spatially distributed bi-stable units that provides a generic model of focal epileptic neuronal system. In the modelled cases of epileptic system with spreading seizures, we found that the removal of the EZ can be predicted by significant decrease of the residual connectivity after performing virtual resection. This decrease commensurate with the increase of the epileptic threshold (the ground truth) . The method also predicts the actual change of connectivity after removing the nodes from the model dynamics. In addition we tested our techniques against the "naïve" virtual resection which is based on simply removing the corresponding nodes from the connectivity measure. The findings in this work can be exploited to increase the efficiency and accuracy of pre-surgical epileptogenic zone localization in cases of focal epileptic seizure onsets.