Beyond the Hospital: A Systematic Review of Wearable Technologies and Machine Learning for Seizure Detection in Epilepsy
Nehashree Bheemu
Advanced Math and Science Academy, Marlborough, USA
Publication date: August 2, 2026
Advanced Math and Science Academy, Marlborough, USA
Publication date: August 2, 2026
DOI: http://doi.org/10.34614/JIYRC2026I48
ABSTRACT
Epilepsy affects millions worldwide, and the unpredictability of seizures can create significant challenges for many patients. Traditional monitoring methods can be impractical for long-term use. Wearable devices combined with machine learning have recently emerged as promising alternatives for continuous, real-world seizure monitoring. This review examines current research on machine learning-based seizure detection using wearable physiological sensors. Across numerous studies, wearable systems demonstrated strong performance in detecting generalized tonic-clonic seizures, with several reporting sensitivities above 90% and low false-alarm rates. However, performance dropped significantly for non-convulsive seizures, which often lack clear motor or autonomic signals. Overall, wearable seizure detection technology shows significant promise but is not yet reliable for all seizure types. Future progress will depend on larger real-world datasets, standardized evaluation protocols, and improved algorithms capable of detecting subtle seizure patterns. With continued development, wearable systems may become valuable tools for improving epilepsy monitoring and patient quality of life.
Epilepsy affects millions worldwide, and the unpredictability of seizures can create significant challenges for many patients. Traditional monitoring methods can be impractical for long-term use. Wearable devices combined with machine learning have recently emerged as promising alternatives for continuous, real-world seizure monitoring. This review examines current research on machine learning-based seizure detection using wearable physiological sensors. Across numerous studies, wearable systems demonstrated strong performance in detecting generalized tonic-clonic seizures, with several reporting sensitivities above 90% and low false-alarm rates. However, performance dropped significantly for non-convulsive seizures, which often lack clear motor or autonomic signals. Overall, wearable seizure detection technology shows significant promise but is not yet reliable for all seizure types. Future progress will depend on larger real-world datasets, standardized evaluation protocols, and improved algorithms capable of detecting subtle seizure patterns. With continued development, wearable systems may become valuable tools for improving epilepsy monitoring and patient quality of life.
