Predictive epileptology: the emergence of a new paradigm for the management of epileptic seizures

OBJECTIVE: To assess the current state and potential use of wearable devices with artificial intelligence (AI) for the prediction of epileptic seizures. MATERIAL AND METHODS: Scientific papers published within the last five years were used for this review, except for fundamental publications. The literature was selected from scientific databases Scopus, PubMed, Google Scholar, and eLibrary. RESULTS: The findings indicate that predicting epileptic seizures with AI-based wearable devices is technologically feasible. Contemporary algorithms, particularly personalized deep learning models, demonstrate predictive accuracy that surpasses random chance by analyzing data from multimodal sensors. Nevertheless, a substantial gap persists between the high accuracy observed in laboratory settings and the more moderate, though clinically meaningful, performance in real-world environments. For patients, these technologies offer significant benefits, including reduced chronic anxiety, enhanced safety, and improved autonomy. Major barriers to widespread adoption include elevated false alarm rates, challenges in predicting non-motor seizures, and considerable regulatory, ethical, and economic constraints. CONCLUSION: Wearable AI-based devices mark a fundamental shift in epilepsy management from reactive response to proactive risk management. While the technology has proven its great potential, it is not yet ready for mass clinical introduction due to existing technical and non-technical barriers. The future of this area lies in the development of integrated «closed-loop» systems that automatically trigger therapeutic interventions based on accurate prognoses. Ultimately, these innovations are not just technological advances, but powerful tools that can improve the quality of life for millions of people.