Epileptic Seizure Detection with an End-to-End Temporal Convolutional Network and Bidirectional Long Short-Term...

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epilepsy diagnosis and treatment rely heavily on accurate seizure detection.
• Electroencephalogram (EEG) monitoring is crucial for understanding seizure activity.
• Existing automatic detection methods face challenges in accuracy and real-time processing.

Purpose of the Study:

• To develop and evaluate a novel end-to-end TCN-BiLSTM model for automatic seizure detection.
• To improve the accuracy and efficiency of epilepsy seizure detection using deep learning.
• To validate the proposed method on established and custom EEG databases.

Main Methods:

• Raw EEG data were filtered using a 0.5-45 Hz band-pass filter.
• A TCN-BiLSTM network was employed for feature extraction and classification of EEG signals.
• Post-processing techniques including moving average filtering, thresholding, and collar technique were applied.

Main Results:

• The TCN-BiLSTM model achieved high performance on the CHB-MIT database: 94.31% sensitivity, 97.13% specificity, and 97.09% accuracy (segment-based).
• Event-based sensitivity reached 96.48% with a low false detection rate (FDR) of 0.38/h on CHB-MIT.
• On the SH-SDU database, segment-based results were 94.99% sensitivity, 93.25% specificity, and 93.27% accuracy, with 99.35% event-based sensitivity and 0.54/h FDR.
• The model processed 1 hour of EEG data in just 5.65 seconds.

Conclusions:

• The proposed TCN-BiLSTM model demonstrates superior performance for automatic seizure detection.
• The method shows significant potential for real-time monitoring and clinical application in epilepsy management.
• The end-to-end deep learning approach offers an efficient and accurate solution for EEG-based seizure detection.