Seizure Prediction Using Wearable EEG

Overview

This study is a non-interventional clinical trial analyzing EEG recordings from people with epilepsy. Participants wear a comfortable EEG headband at home for several weeks. The goal is to study changes in brain activity that occur before seizures (called "pre-ictal patterns") and to test whether a software algorithm can predict seizures in real-time based on these patterns. No treatments or medications are being tested. The study will help evaluate whether seizure prediction is possible using wearable EEG devices and can support the development of future tools that give patients early warnings before seizures occur.

This observational study aims to evaluate the feasibility of real-time seizure prediction using non-invasive, wearable EEG devices in patients with epilepsy. The study focuses on identifying pre-ictal EEG patterns-subtle changes in brain activity that occur prior to seizure onset-and validating a prediction algorithm based on these patterns. Epileptic seizures often occur unpredictably, significantly affecting patients' quality of life and safety. Existing seizure detection systems operate only after seizure onset. In contrast, predicting seizures before they occur could enable timely interventions, increase patient autonomy, and reduce the risks associated with uncontrolled seizures. The study involves home use of consumer-grade wearable EEG devices (e.g., BrainBit and Muse headbands), which transmit EEG data via Bluetooth to a mobile app developed by the sponsor. Participants are instructed to wear the device daily for at least 12 weeks. The mobile app provides feedback on signal quality and securely uploads the data to the cloud for analysis. Participants can record seizures through the app, and researchers will also collect medical records for additional clinical annotations when available. The prediction algorithm being tested uses personalized calibration and advanced statistical control of false alarm rates to ensure clinical viability. The algorithm was initially developed and tested using retrospective hospital-grade EEG data and publicly available datasets. This trial extends that work into the real world, evaluating the algorithm's performance prospectively on wearable data. Key aims include: Evaluating the usability of wearable EEG devices for long-term home use in a diverse patient population. Identifying consistent pre-ictal EEG features within and across patients. Validating the performance of the seizure prediction algorithm in terms of sensitivity, specificity, and false alarm rate. Exploring the consistency of pre-ictal patterns across multiple seizures for the same patient. This feasibility trial is non-interventional and does not alter participants' treatment plans. All data are collected passively and analyzed after being de-identified. Ethics approvals were obtained. The study is expected to contribute critical evidence toward the development of a clinically useful, AI-powered seizure forecasting system for real-world use.