Seizures detection using multimodal signals: a scoping review.

Area of Science:

• Neurology
• Biomedical Engineering
• Digital Health

Background:

• Epileptic seizures affect 65 million globally, with 30% experiencing refractory epilepsy unresponsive to medication.
• Unpredictable seizures necessitate continuous monitoring systems for patient safety and treatment.
• Autonomic changes during seizures are key indicators for detection and prediction research.

Purpose of the Study:

• To review current research on portable, wearable devices for seizure detection and prediction.
• To inform future directions in continuous, ambulatory seizure tracking.
• To assess the feasibility of non-cerebral physiological signals for seizure monitoring.

Main Methods:

• A scoping review methodology following PRISMA guidelines.
• Systematic literature search conducted on PubMed and IEEE databases.
• Analysis of 30 selected articles focusing on wearable device implementation for seizure detection.

Main Results:

• Most studies utilized offline analysis and consumer-grade wearable devices.
• Autonomic channel monitoring (ACM) was the dominant modality, with algorithms like SVM and Random Forest widely used.
• Single modality sensitivity ranged from 33.2% to 100% (FAR 0.096–14.8 d⁻¹), while multimodality achieved 51%–100% sensitivity (FAR 0.12–17.7 d⁻¹).

Conclusions:

• Seizure detection systems using non-cerebral physiological signals show promising performance.
• These systems are particularly effective for detecting motor seizures and those with significant autonomic changes.
• Further development in wearable technology can enhance continuous seizure monitoring in ambulatory settings.