Hybrid Ensemble Learning for EEG-Based Epileptic Seizure Forecasting

2026-09-26

AI Summary

This study addresses the challenges of limited cross-patient generalization and uncontrolled false alarm rates in epileptic seizure prediction by proposing a probability-calibrated hybrid stacking ensemble framework. The method integrates five deep learning and three classical machine learning models, employing logistic regression as a meta-learner for predictive calibration. A comprehensive pipeline is constructed incorporating EEG preprocessing, feature extraction, SMOTE oversampling, and Platt scaling, with rigorous leave-one-patient-out (LOPO) cross-validation ensuring unbiased evaluation. Evaluated on the CHB-MIT dataset, the proposed framework achieves 74.2% seizure-level sensitivity, a false alarm rate of 1.24 per hour, and an average prediction horizon of 16.9 minutes. These results significantly outperform the Oracle baseline, demonstrating enhanced cross-patient generalization performance suitable for clinical deployment.

Abstract

Epileptic seizure forecasting aims to provide actionable warnings before seizure onset, yet patient-independent generalization and false-alarm control remain major challenges. We propose a calibrated hybrid ensemble for EEG-based seizure forecasting that combines five deep learning models and three classical machine learning models through a logistic regression stacking meta-learner. The proposed pipeline integrates signal preprocessing, handcrafted feature extraction, class-imbalance handling, probability calibration, and clinically motivated post-processing. We evaluate the framework on CHB-MIT using strict Leave-One-Patient-Out (LOPO) cross-validation, with threshold and post-processing parameters selected only on held-out meta data. On the filtered cohort, excluding patients with anomalous preictal rates below 1\% or above 15\%, the model achieves 74.2\% seizure-level sensitivity at 1.24 false alarms per hour, with an average warning time of 16.9 minutes. A test-tuned oracle constrained to the target false-alarm budget achieves 60.9\% sensitivity at 0.951 false alarms per hour, highlighting the importance of reporting sensitivity together with realized false-alarm rates. Our code is available at: https://github.com/DanaMason/IEEE-CARS-Hybrid-Ensemble-Learning-for-EEG-Based-Epileptic-Seizure-Forecasting

Problem

• Epileptic seizure forecasting
• EEG
• Patient-independent generalization
• False-alarm control

Innovation

• Hybrid Ensemble Learning
• Seizure Forecasting
• Stacking Meta-learner
• Probability Calibration
• LOPO Cross-validation