Related Experiment Video
Updated: Jul 17, 2025
05:58
Application of an Amplitude-integrated EEG Monitor Cerebral Function Monitor to Neonates
Published on: September 6, 2017
38.9K
Automated diagnosis of EEG abnormalities with different classification techniques
Essam Abdellatef1, Heba M Emara2, Mohamed R Shoaib3
1Department of Electronics and Communications, Delta Higher Institute for Engineering and Technology (DHIET), 35511, Mansoura, Egypt.
Medical & Biological Engineering & Computing
|September 6, 2023
Summary
This study introduces two advanced methods for automatic seizure detection and prediction from EEG data. Both machine learning and deep learning approaches achieved high accuracy, outperforming existing techniques.
Area of Science:
- Biomedical Engineering
- Computational Neuroscience
- Signal Processing
Background:
- Epileptic seizure detection and prediction from Electroencephalograms (EEGs) are complex due to low signal-to-noise ratios, patient-specific seizure variations, and data limitations.
- Existing methods face challenges in accurately identifying and forecasting seizures in clinical settings.
Purpose of the Study:
- To develop and evaluate two novel approaches for enhanced automatic seizure detection and prediction using EEG signals.
- To overcome the limitations of current methods by employing advanced signal processing and machine learning techniques.
Main Methods:
- Developed a Machine Learning (ML) model utilizing features from the Hilbert Marginal Spectrum (HMS) domain, including entropy, higher-order statistics, and sub-band energies, classified with SVM, LR, and KNN.
- Implemented a deep learning approach using Convolutional Neural Networks (CNNs) with residual learning on EEG spectrograms for seizure detection, prediction, and 3-state classification.
Main Results:
- The HMS-based ML model achieved 100% accuracy for seizure detection and prediction.
- The CNN-based model demonstrated high accuracies: 97.66% (Seizure vs. Pre-Seizure), 95.59% (Non-Seizure vs. Seizure), and 94.51% (Non-Seizure vs. Seizure vs. Pre-Seizure).
- Both methods outperformed state-of-the-art techniques on the CHB-MIT EEG dataset.
Conclusions:
- The proposed HMS-based and CNN-based approaches are highly effective for automatic epileptic seizure detection and prediction.
- These methods offer significant improvements over existing techniques, paving the way for more reliable seizure monitoring.
- The study highlights the potential of advanced computational methods in improving epilepsy management.
