Epilepsy-IEDs: An automated machine learning model for detecting interictal epileptiform discharges from scalp electroencephalograms.
Journal:
iScience
Published Date:
Jun 22, 2026
Abstract
Interictal epileptiform discharges (IEDs) are essential for epilepsy diagnosis, yet visual electroencephalogram (EEG) analysis remains subjective and laborious. To address this, we developed Epilepsy-IEDs, an automated machine learning model for IED detection. Trained on 141 scalp EEG recordings (2,597 IEDs and 4,633 non-IEDs), the model was evaluated using four algorithms, with a separate daytime analysis to control for sleep effects. The Extreme Gradient Boosting (XGBoost)-based model achieved the highest performance, with sensitivities of 84.6% (area under the curve [AUC] = 0.966) and 87.1% (AUC = 0.973) on the full and daytime datasets, respectively, and demonstrated robust generalization on held-out epilepsy patients (AUC = 0.878-0.890), with a specificity of 71.54% on a non-epilepsy cohort. A simplified 10-feature variant maintained strong performance (AUC = 0.959). The Epilepsy-IEDs model provides an accurate, interpretable tool for IED detection, with a streamlined version suitable for integration into clinical EEG workflows.
Authors
Keywords
No keywords available for this article.