Predicting Epileptic Seizures Using EfficientNet-B0 and SVMs: A Deep Learning Methodology for EEG Analysis.

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Epilepsy management presents challenges in timely intervention.
• Accurate seizure prediction can significantly improve patient outcomes.
• Existing methods require enhanced robustness and adaptability.

Purpose of the Study:

• To develop a novel and robust framework for seizure prediction.
• To combine deep learning (EfficientNet-B0 CNN) with ensemble machine learning (SVMs).
• To leverage spectral and spatial features from EEG signals for improved prediction.

Main Methods:

• Utilized a convolutional neural network (CNN) based on EfficientNet-B0.
• Employed an ensemble of six Support Vector Machines (SVMs) with a voting mechanism.
• Extracted normalized Short-Time Fourier Transform (STFT) and channel correlation features from EEG signals.

Main Results:

• Achieved high accuracies (96.12% for 10 min, 94.89% for 20 min, 94.21% for 30 min).
• Demonstrated high sensitivities (95.21% for 10 min, 93.98% for 20 min, 93.55% for 30 min).
• Validated on the CHB-MIT EEG dataset, showing robustness and adaptability compared to state-of-the-art methods.

Conclusions:

• The proposed framework offers a robust and computationally efficient solution for seizure prediction.
• The combination of EfficientNet-B0 and SVM ensemble enhances prediction reliability.
• This approach holds significant potential for improving epilepsy management through timely interventions.