Real-Time Epileptic Seizure Detection Based on Deep Learning

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epilepsy is a common neurological disorder requiring prolonged video electroencephalogram (EEG) monitoring for diagnosis.
• Current video EEG analysis places a significant burden on both escorts and medical professionals due to lengthy monitoring periods and large data volumes.

Purpose of the Study:

• To develop and evaluate a deep neural network for the real-time detection of epileptic seizures from EEG data.
• To reduce the burden on caregivers and improve the efficiency of EEG data analysis for epilepsy diagnosis.

Main Methods:

• A deep neural network model was designed and optimized using specified signal representations for real-time seizure detection.
• Patient-specific neural networks were trained using EEG data from the CHB-MIT public dataset.
• A smoothing filter was applied to the model's output to enhance performance.

Main Results:

• The developed model achieved a recall of 97% for real-time seizure event detection.
• The system demonstrated a low false alarm rate of 0.219 per hour.
• The detection latency was measured at an average of 3.4 seconds.

Conclusions:

• The proposed deep neural network method enables accurate and efficient real-time detection of epileptic seizures.
• This technology holds significant potential for integration into clinical systems to streamline epilepsy diagnosis and patient care.