Wearable Artificial Intelligence for Epilepsy: Scoping Review.

Area of Science:

• Medical Devices and Wearable Technology
• Artificial Intelligence in Healthcare
• Neurology and Epilepsy Management

Background:

• Epilepsy affects 50 million globally, necessitating continuous monitoring for effective management.
• Wearable AI technologies leverage physiological signals and machine learning for seizure detection and prediction.
• A comprehensive overview of advances and challenges in wearable AI for epilepsy is needed.

Purpose of the Study:

• To comprehensively review and map the literature on AI-driven wearable technologies for epilepsy.
• To identify device characteristics, AI methodologies, biosignal measurements, and validation approaches.
• To pinpoint research gaps in the field of wearable AI for epilepsy management.

Main Methods:

• A scoping review was conducted following PRISMA-ScR guidelines.
• Systematic search across six databases (Scopus, MEDLINE, Embase, ACM, IEEE, Google Scholar) up to December 2023.
• Included studies on AI algorithms for epilepsy using noninvasive wearable devices; excluded nonwearable or in-body devices.

Main Results:

• 67 studies met inclusion criteria, with significant research growth since 2021.
• Activity, cardiovascular, brain activity, and skin conductance were primary biosignals monitored.
• Support vector machines, random forests, and CNNs were common AI models; focus on seizure detection over prediction.
• Limited clinical validation and predominance of closed-source data were noted.
• Leave-one-out and k-fold cross-validation were common, with video-EEG as the primary reference.

Conclusions:

• Wearable AI shows promise for real-time epilepsy monitoring and early seizure detection.
• Standardization of validation methods and open data exchange are crucial for clinical impact.
• Development of energy-efficient algorithms is needed for real-world deployment of wearable devices.