(662) Ceribell Point-of-Care EEG Provides High-Quality Signal for up to 24 Hours of Continuous Recording

Introduction: Point-of-care EEG systems can be set-up rapidly on patients at risk for seizures in acute care settings. They can also be used for continuous monitoring, especially in hospitals with limited access to conventional EEG. Machine learning algorithms can use data from these systems to continuously monitor seizures, but their performance can be affected by recording quality. Here, we tested whether the duration of point-of-care EEG monitoring using Ceribell is associated with a decrease in signal quality that can impact the performance of Clarity, an algorithm for automated seizure detection.

Methods: The Clarity algorithm divides the EEG into 10 seconds segments and automatically discards those for which the impedance of the electrodes, range or root mean square of the signal are too high. We retrospectively analyzed a dataset of 336 point-of-care EEGs lasting between 4 and 24 hours collected in the emergency department and intensive care units of 4 hospitals. We used Wilcoxon signed-rank tests to determine whether the number of segments discarded in the first hour of EEG was significantly different from the number of segments discarded in each hour after the first.

Results: The median duration of the EEG recordings was 7.53 hours, 153 recordings were longer than 8 hours and 86 recordings were longer than 12 hours. The number of segments discarded was not significantly different between the first hour of EEG and each of the subsequent hours, up to 24 hours (all p>0.05). The median percentage of segments discarded was < 5% at every hour of recording.

Conclusions: The number of EEG segments discarded by Clarity because of low signal quality did not increase with duration of the recording and remained low up to 24 hours. Our results show that Ceribell point-of-care EEG can be used for long continuous monitoring and for long continuous automated seizure detection using Clarity.