부산대학교 도서관

Objective: Localization and mapping of seizure-generating brain tissue, i.e., seizure onset zone (SOZ) is essential to ensure an excellent patient outcome after surgical resection. The clinical approach is to record spontaneous seizures with intracranial EEG (iEEG) and determine SOZ. However, this practice is burdened by inter-patient variability, temporal variability, time-consuming data annotation, and long and variable waiting period for seizures to happen. Approach: Here, we use data from intracranial monitoring of 28 patients with neocortical focal epilepsy. Kurtosis, complexity, activity, mobility, mean, median, min, max, peak to peak, variance, standard deviation, root mean square, and interquartile were extracted as features from the time domain in two frequency bands (12–55 Hz and 55–80 Hz). The features were extracted from segments of inter-ictal iEEG from 8962 channels and tested by Wilcoxon rank sum test with Bonferroni correction of alpha to compare if mean of the feature differs in SOZ versus non-SOZ in each patient individually. Results: From all features, kurtosis, maximum, minimum, peak to peak, standard deviation, root mean square, variance, interquartile shown consistent differences between SOZ and non-SOZ channels across patients (p