Deep learning-based real-time seizure detection and multi-seizure classification on pediatric EEG.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epilepsy is a common neurological disorder in children, necessitating accurate and timely detection methods.
• Current seizure detection methods may lack the accuracy and real-time capabilities required for effective pediatric epilepsy management.

Purpose of the Study:

• To develop and validate a deep learning-based system for real-time detection and classification of multiple seizure types in pediatric patients.
• To assess the performance of deep learning models in analyzing electroencephalography (EEG) data for clinical application.

Main Methods:

• Retrospective collection of EEG recordings from pediatric patients (3 months to 18 years) diagnosed with various epilepsy types.
• Downsampling of EEG data to 200 Hz for real-time processing and application of deep learning models (ResNet with Long-Short Term Network, ResNet50).

Main Results:

• The ResNet with Long-Short Term Network achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.98 and an Area Under the Precision-Recall Curve (APROC) of 0.73 for real-time seizure detection.
• ResNet50 demonstrated superior performance in multi-class seizure detection, with an AUROC of 0.99 and an Area Under the Precision-Recall Curve (APPRC) of 0.99.

Conclusions:

• The proposed deep learning approach provides robust, real-time detection and classification of multiple seizure types in pediatric epilepsy.
• The system demonstrates effective application to real-world clinical EEG datasets, offering realistic performance and speed for monitoring childhood seizures.