Personalizing Seizure Detection for Individual Patients by Optimal Selection of EEG Signals

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Electroencephalography (EEG) is crucial for diagnosing neurological disorders like epilepsy.
• Current 21-electrode EEG systems can be cumbersome; personalized channel selection offers improved efficiency.
• Wearable devices require compact, low-power solutions for continuous monitoring.

Purpose of the Study:

• To develop a lightweight deep learning system for real-time seizure detection.
• To implement a patient-specific, two-channel selection method for enhanced EEG analysis.
• To create a reliable and interpretable seizure detection system for personalized wearable devices.

Main Methods:

• An efficient Convolutional Neural Network (CNN) was designed for processing EEG data.
• A data-driven mechanism automatically selects the two most informative EEG channels per patient.
• The system was validated on the CHB-MIT EEG dataset.

Main Results:

• The two-channel system achieved an average balanced accuracy of 0.83.
• A low false-positive rate of approximately 0.1 per hour was recorded.
• Performance matched or exceeded state-of-the-art systems using more channels, with a 30% reduction in false positives.

Conclusions:

• Patient-specific, two-channel EEG analysis is effective for reliable seizure detection.
• The developed deep learning system is suitable for compact, efficient wearable epilepsy monitoring.
• This interpretable approach facilitates personalized neurological disorder management.