1157 Co-Localization of Epileptiform Discharges With Preoperative Diffusion-Weighted Imaging Biomarkers in Pediatric Epilepsy

INTRODUCTION: Minimally invasive procedures such as MRI-guided LiTT are increasingly being utilized in pediatric epilepsy patients; however, their clinical efficacy has been limited by challenges in precisely localizing epileptogenic tissues. A precise and accurate means to highlight epileptogenic tissues noninvasively will augment the utility of LiTT. Previous studies have suggested that an apparent diffusion coefficient (ADC) map may highlight microstructural abnormalities commonly seen in temporal lobe epilepsy and provide a prognosis for seizure freedom following LiTT. METHODS: Twelve pediatric epilepsy patients who underwent unilateral amygdalohippocampal SEEG were reviewed. Based on preoperative ADC sequences of each patient, a machine-learning algorithm was employed to detect hyperintensive ADC clusters within the hippocampus and amygdala ipsilateral to SEEG-implanted hemisphere. In parallel, coordinates of SEEG-electrode contacts within 30mm of each ADC cluster where abnormal interictal epileptiform discharges (“active” contacts), and contacts with no discharges (“passive” contacts) were identified. Euclidian distances between centroid of ADC hyperintensive clusters and locations of SEEG electrode contacts for each patient were calculated and compared between active and passive contacts. RESULTS: Cluster-SEEG contact distance was relatively shorter with active contacts than passive contacts in the hippocampus (17.3mm vs. 21.7mm, p=0.02). After fitting a multivariate logistic regression model predicting whether a contact is active or passive from cluster-contact distance measures, ROC analysis yielded a significant predictive capacity for hippocampal clusters (AUC=0.95, p=0.0013) for mean cluster distance, and both hippocampal (AUC=0.72, p <0.0001) and amygdala clusters (AUC=0.72, p=0.012) for individual cluster distance. CONCLUSIONS: Our results suggest that ADC hyperintensive clusters may co-localize with epileptiform activity detected in SEEG contacts, especially in hippocampus. Furthermore, our findings highlight the potential utility of preoperative diffusion-weighted imaging in precisely capturing locations with epileptiform discharges inferred by SEEG recordings.