Abstracts

Rationale: Continuous EEG (cEEG) monitoring in an Epilepsy Monitoring Unit (EMU) is the gold standard diagnostic procedure for patients with medically refractory seizures and also critical for epilepsy presurgical evaluation. However, the burden of high volumes of data and delayed recognition of electrographic seizures continue to be the major challenges in offering cEEG monitoring at many clinical centers. The goal of this study was to validate the performance of an all-in-one EEG and seizure monitoring handheld device, CereScope, which is designed to acquire a high quality, full 10-20 system scalp EEG recording and to provide real-time seizure alerts to designated caregivers through email notifications. Specifically, the study tested the following hypotheses: (1) EEG signals acquired from CereScope have comparable quality to those recorded from clinical EEG systems; (2) Real-time event detections match those obtained from post hoc analyses; (3) Email alerts of possible seizure events deliver within 2 minutes of the detection time; and (4) CereScope’s real-time seizure detection performance has at least 90% sensitivity and a false detection rate less than 3 per 24 hours of recording.  We envision that such a device can be clinically useful for inpatient EMU, ambulatory, and ICU patients undergoing cEEG monitoring. Methods: A total of 100 EMU patients from the four collaborating clinical centers (AGH, Emory, MUSC, and Vanderbilt) were enrolled in this study. Each patient underwent simultaneous scalp cEEG recordings (CereScope and Natus/Nihon-Kohden EEG system) using a specially designed configuration with signal split cables. To establish a “gold standard” for performance assessment, EEG segments from 57 subjects, which included all patients with seizures and randomly selected patients without seizures based on their clinical EEG reports, were randomly sampled and reviewed by three independent EEG experts to determine the signal quality as well as the occurrences of electrographic seizures. In addition, each EEG recording was processed post hoc entirely and the resulting detections were compared with the real-time email alerts to assess the reliability of the real-time process. The mean alert delay was estimated based on the time difference between the detection time and when the alert email was delivered from the mail server. Results: (1) Based on the independent expert review, in 95% of the sampled EEG segments, CereScope's recording was comparable or better than those acquired through the clinical systems (95% CI = [0.88, 0.99]). (2) The detection results from the real-time process were highly similar to those from the post hoc analysis (44 "real-time" detections vs. 42 "offline" detections); certain mismatches were expected due to the difference in the continuity of EEG files. (3) The mean delay between the actual seizure detection and the completion of sending an email alert was approximately 94 seconds (shorter than 2 minutes, p-value < 0.05). (4) CereScope's seizure detection software correctly detected 28 (out of 29) electrographic seizures (96.6%) and had a false detection rate of 0.34 per 24 hours of EEG recording. Conclusions: The results suggest that CereScope, a handheld EEG device, can operate the real-time process of its sophisticated signal processing algorithm for detecting ictal EEG patterns and sending event alerts without affecting the data acquisition performance of the system, and vice versa. These observations support our hypothesis that CereScope’s simultaneous multi-task processes can acquire high-quality EEG recordings and provide accurate real-time seizure alerts during continuous multi-day EEG monitoring. Funding: NIH/NINDS SBIR Grant Number: 2R44NS064647