Interictal network synchrony and local heterogeneity predict epilepsy surgery outcome among pediatric patients

Abstract

Objective: Epilepsy is a disorder of aberrant cortical networks. Researchers have proposed that characterizing presurgical network connectivity may improve the surgical management of intractable seizures, but few studies have rigorously examined the relationship between network activity and surgical outcome. In this study, we assessed whether local and global measures of network activity differentiated patients with favorable (seizure-free) versus unfavorable (seizure-persistent) surgical outcomes. Methods: Seventeen pediatric intracranial electroencephalography (IEEG) patients were retrospectively examined. For each patient, 1,200 random interictal epochs of 1-s duration were analyzed. Functional connectivity networks were constructed using an amplitude-based correlation technique (Spearman correlation). Global network synchrony was computed as the average pairwise connectivity strength. Local signal heterogeneity was defined for each channel as the variability of EEG amplitude (root mean square) and absolute delta power (μV2/Hz) across epochs. A support vector machine learning algorithm used global and local measures to classify patients by surgical outcome. Classification was assessed using the Leave-One-Out (LOO) permutation test. Results: Global synchrony was increased in the seizure-persistent group compared to seizure-free patients (Student's t-test, p = 0.006). Seizure-onset zone (SOZ) electrodes exhibited increased signal heterogeneity compared to non-SOZ electrodes, primarily in seizure-persistent patients. Global synchrony and local heterogeneity measures were used to accurately classify 16 (94.1%) of 17 patients by surgical outcome (LOO test, iterations = 10,000, p < 0.001). Significance: Measures of global network synchrony and local signal heterogeneity represent promising biomarkers for assessing patient candidacy in pediatric epilepsy surgery.