Classification of IED-free EEG Responses for Assisted Epilepsy Diagnosis

2026-05-19

AI Summary

This study addresses the diagnostic challenge of epilepsy in routine electroencephalography (EEG) recordings lacking interictal epileptiform discharges (IEDs), where current activation procedures—such as photic stimulation and hyperventilation—are limited by subjective interpretation. The work systematically demonstrates, for the first time, that stimulus-evoked EEG responses—particularly those elicited by intermittent photic stimulation—contain discriminative information even in the absence of IEDs. A reproducible machine learning pipeline is developed, integrating multi-domain features from time, frequency, wavelet, and functional connectivity domains, combined with a stacking ensemble and leave-one-subject-out cross-validation. Evaluated on the TUH dataset, the model achieves AUCs of 97.8% (resting state) and 94.1% (photic stimulation); on an independent EMC clinical cohort, it attains 79.4% AUC under photic stimulation, while hyperventilation performance improves markedly after reactivity stratification, confirming that multi-domain feature fusion enhances model robustness and clinical applicability.

Abstract

Diagnosing epilepsy is challenging when routine EEGs lack interictal epileptiform discharges (IEDs). Intermittent photic stimulation (IPS) and hyperventilation (HV) can increase diagnostic yield, but their interpretation is subjective. We propose a reproducible pipeline that classifies EEG recordings acquired during stimulation procedures, using machine-learning features spanning temporal, spectral, wavelet, and connectivity domains, and a stacked ensemble to combine complementary feature sets. Performance is evaluated with leave-one-subject-out (LOSO) cross-validation on the TUH Epilepsy Corpus and a clinical Erasmus MC (EMC) cohort, including IED-free analyses on TUH. On TUH, ensembles achieve up to 97.8\% AUC / 93.1\% BAC on IED-free resting-state EEG and 94.1\% AUC / 86.8\% BAC on IED-free IPS. On EMC, IPS provides the strongest discrimination (79.4\% AUC / 73.9\% BAC), while HV performance benefits from stratifying subjects by responsiveness. These results indicate that stimulation-evoked activity, particularly IPS, contains meaningful discriminative information for IED-free epilepsy classification and that multi-domain ensembling improves robustness.

Problem

• epilepsy diagnosis
• IED-free EEG
• intermittent photic stimulation
• hyperventilation
• EEG classification

Innovation

• multi-domain features
• stacked ensemble
• IED-free EEG
• intermittent photic stimulation
• epilepsy classification