A Multi-View Deep Learning Framework for EEG Seizure Detection.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epilepsy management requires accurate seizure detection from electroencephalogram (EEG) signals.
• Manual analysis of long-term EEG data is time-consuming and prone to errors.
• Pervasive sensing technologies enable continuous monitoring of epilepsy patients.

Purpose of the Study:

• To develop an automated, accurate, and efficient system for detecting epileptic seizures using multi-channel scalp EEG signals.
• To propose a unified multi-view deep learning framework for capturing brain abnormalities related to seizures.
• To enhance seizure detection by effectively utilizing inter and intra-channel correlations in EEG data.

Main Methods:

• A unified multi-view deep learning framework was designed, integrating unsupervised EEG reconstruction and supervised seizure detection.
• An autoencoder-based multi-view learning model was constructed to capture both inter and intra-channel EEG correlations.
• A channel-aware seizure detection module with a channel-wise competition mechanism was incorporated to focus on relevant EEG channels.

Main Results:

• The proposed framework achieved high performance in EEG seizure detection.
• Achieved an average accuracy of 94.37% and an F1-score of 85.34% using 5-fold subject-independent cross-validation.
• Outperformed nine traditional and deep learning baseline methods on a benchmark scalp EEG dataset.

Conclusions:

• The developed multi-view deep learning framework is a powerful and effective method for automated EEG seizure detection.
• The channel-aware mechanism enhances the model's ability to focus on critical EEG channels for improved seizure identification.
• This approach offers a promising solution for reducing the burden of manual EEG analysis in epilepsy monitoring.