Epileptic seizure forecasting with wearable-based nocturnal sleep features

Area of Science:

• Biomedical Engineering
• Neuroscience
• Wearable Technology

Background:

• Non-invasive biomarkers are crucial for advancing seizure forecasting in epilepsy.
• Wearable devices offer a promising avenue for continuous physiological monitoring.

Purpose of the Study:

• To develop and evaluate a seizure-day forecasting algorithm using nocturnal sleep features from a smart shirt.
• To assess the efficacy of a smart shirt-based system for predicting seizures in individuals with epilepsy.

Main Methods:

• Seventy-eight epilepsy patients wore a biometric smart shirt measuring ECG, respiration, and accelerometry.
• Ten sleep features were extracted and normalized; a support vector machine classifier was trained for 16- and 24-hour seizure forecasting.
• Performance was evaluated using a nested leave-one-patient-out cross-validation.

Main Results:

• The algorithm achieved improvement over chance (IoC) in 48.7% (16-h) and 40% (24-h) of patients.
• For successful forecasts, mean IoC was 34.3% (16-h) and 34.2% (24-h), with sensitivities of 86.0% (16-h) and 64.4% (24-h).

Conclusions:

• Smart shirt-based sleep analysis shows potential for non-invasive seizure-day forecasting in a subset of epilepsy patients.
• Further research in long-term, residential settings is needed to develop practical seizure forecasting devices.