Near-Real-Time Epileptic Seizure Detection with Reduced EEG Electrodes: A BiLSTM-Wavelet Approach on the EPILEPSIAE...

Area of Science:

• Neurology
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy is a chronic neurological disorder marked by recurrent seizures due to abnormal brain activity.
• Near-real-time seizure detection is crucial for patient safety, timely intervention, and enhancing quality of life, especially in drug-resistant epilepsy.
• Current detection methods often require extensive monitoring and may impact patient comfort.

Purpose of the Study:

• To develop and validate a low-latency framework for near-real-time epileptic seizure detection.
• To improve clinical reliability and patient comfort by reducing the number of required Electroencephalogram (EEG) electrodes.
• To assess the performance of a Bidirectional Long Short-Term Memory (BiLSTM) network combined with wavelet-based features.

Main Methods:

• Utilized the EPILEPSIAE dataset, analyzing EEG signals from 161 patients with 1032 recorded seizures.
• Integrated BiLSTM networks with wavelet-based feature extraction, combining raw EEG data with extracted features for enhanced signal representation.
• Conducted electrode reduction experiments to identify the minimum number of electrodes necessary for maintaining detection performance.

Main Results:

• The optimized BiLSTM model demonstrated 86.9% accuracy and 86.1% recall.
• Achieved a low average detection delay of 1.05 seconds and a processing time of 0.065 seconds per 0.5-second EEG window.
• Established that reliable seizure detection is possible with as few as six strategically placed electrodes, comparable to full electrode configurations.

Conclusions:

• The proposed BiLSTM-wavelet approach offers a clinically viable and computationally efficient solution for near-real-time epileptic seizure detection.
• The framework is wearable-friendly due to the reduced number of EEG channels required.
• This method enhances the practicality and accessibility of seizure detection systems for patients with epilepsy.