Internal and External Validation of Comprehensive High-Frequency Activity Biomarkers for Epilepsy Surgery

Area of Science:

• Neuroscience
• Epileptology
• Biomarker Discovery

Background:

• Epilepsy surgery relies on accurate identification of the seizure onset zone (SOZ).
• Intracranial EEG (iEEG) biomarkers, particularly high-frequency activity (HFA), are proposed for SOZ localization and outcome prediction.
• Generalizability of these biomarkers across diverse patient populations and etiologies remains under-validated.

Purpose of the Study:

• To develop and validate iEEG-derived biomarkers for SOZ localization and postoperative seizure outcome prediction in drug-resistant focal epilepsy.
• To assess the generalizability of these biomarkers across independent, multi-center, and multi-etiology cohorts.
• To investigate the influence of epilepsy etiology on the predictive performance of biomarkers.

Main Methods:

• Retrospective analysis of iEEG data from 233 patients with drug-resistant focal epilepsy across multiple centers.
• Quantification of HFA occurrence rates and morphological features.
• Development of predictive models for SOZ localization and postoperative seizure outcome, validated on independent external cohorts.

Main Results:

• HFA rate, spectral entropy, and power were key features for SOZ classification, achieving AUCs up to 0.86 in external cohorts.
• SOZ localization models demonstrated cross-center generalizability.
• Postoperative seizure outcome prediction was reliable in the derivation cohort (AUC up to 0.70) but showed significant decline in external cohorts, particularly for encephalomalacia (AUC ≤ 0.46).

Conclusions:

• Integration of HFA and morphological features provides a generalizable biomarker for SOZ localization.
• Postoperative outcome prediction is highly dependent on epilepsy etiology, with limited generalizability.
• Surgical strategies targeting HFA-rich areas may be ineffective or detrimental in patients with encephalomalacia.