426 Localization of Seizure Onset Zones Using Deep Learning on Resting-State Stereoelectroencephalography

View at publisher Summarise with VIRAT
Authors: Sameer C. Sundrani, Graham Walter Johnson, Derek J. Doss, Ghassan S. Makhoul, Bruno Hidalgo Monroy Lerma, Anas J. Reda, Addison C. Cavender, Emily Liao, Baxter Rogers, Shawniqua Williams Roberson, Sarah K. B. Bick, Victoria L. Morgan
+1 more Dario J. Englot
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

INTRODUCTION: Epilepsy affects more than 50 million people worldwide, and one-third of patients continue to have seizures despite maximal medical management. Surgical resection may be curative if patients’ seizures localize to a discrete brain region, termed a seizure onset zone. Localization is often confirmed with stereotactic electroencephalography; however, this may require patients to stay in the hospital for weeks to capture spontaneous seizures. METHODS: Using more than 1,000,000 interictal stereotactic electroencephalography segments collected from 78 patients, we evaluated a multichannel, multiscale 1-dimensional convolutional neural network via five-fold cross-validation and testing to classify seizure onset zones. RESULTS: Our models achieved a seizure onset zone classification sensitivity of 0.702 (95% CI: [0.549, 0.805]), specificity of 0.741 (95% CI: [0.652, 0.835]), and accuracy of 0.738 (95% CI: [0.687, 0.795]), which was significantly better than models trained on random labels on held-out test sets. The models performed well across the entire brain, with top five region performance demonstrating accuracies between 70.0 to 88.4%. When split by outcomes, the models performed significantly better on patients with favorable Engel outcomes after resection or who were RNS responders. Finally, SHapley Additive exPlanation (SHAP) value analysis on median normalized input data demonstrated notably high feature importance to interictal spikes and large deflections, while similar analyses on histogram equalized data revealed previously unobserved differences in feature importance assignments to low-amplitude segments. CONCLUSIONS: This work serves as evidence that deep learning on short segments of resting-state stereotactic electroencephalographic data can localize seizure onset zones across the brain. Our results corroborate modern understandings of interictal epileptiform discharges and may be used to uncover novel interictal morphologies. Finally, application of our models may reduce dependence on observing clinical seizures for localization and shorten pre-surgical evaluation time for drug-resistant epilepsy patients.

Fields of Study & Subjects

Epilepsy research and treatment EEG and Brain-Computer Interfaces Machine Learning in Healthcare

Publication details

Journal / source
Neurosurgery
Volume
72
Issue
Supplement_1
Pages
121–122
Publication date
26 March 2026
Language
EN