An edge-AI enabled wearable platform for real-time epileptic seizure detection with geolocated alerting

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Epilepsy poses a significant global health challenge, especially in areas lacking continuous medical monitoring.
• Existing monitoring solutions are often inaccessible or impractical in resource-limited environments.

Purpose of the Study:

• To develop and evaluate a wearable, edge-based AI system for real-time epileptic seizure detection.
• To enhance detection reliability through multi-sensor fusion and on-device machine learning.
• To provide immediate alerts with location data for timely medical intervention.

Main Methods:

• Integration of accelerometers, photoplethysmography (PPG), and GPS sensors in a wearable device.
• Deployment and assessment of machine learning models (SVM, NN, Random Forest) directly on the edge device.
• Utilizing sensor fusion to improve the accuracy and specificity of seizure detection.
• Implementation of an SMS alert system for caregivers upon seizure detection.

Main Results:

• The optimized random forest algorithm demonstrated the highest accuracy and fastest response time among tested models.
• Sensor fusion significantly improved detection specificity and maintained a low false alarm rate.
• The system successfully generated real-time SMS alerts with precise location data upon simulated seizure detection.

Conclusions:

• The developed wearable AI system is a cost-effective, self-contained solution for real-time epilepsy seizure detection.
• The platform has strong potential to enhance patient safety and facilitate rapid response in low-resource settings.
• This technology can bridge the gap in continuous medical monitoring for epilepsy patients in underserved regions.