Abstracts

Rationale: Clarity, an automated seizure detection algorithm, uses point-of-care electroencephalogram (POC EEG) to provide continuous seizure burden (SzB) monitoring and bed-side alerts for suspected status epilepticus (SE). Since its release in 2020, Clarity has been regularly updated to improve performance in diverse seizure types. In this study, we evaluated the performance of Clarity’s latest version in a large retrospective labeled dataset.

Methods: We analyzed a retrospective dataset containing 1148 POC EEGs collected from more than 10 hospitals. All epochs from each EEG recording were annotated by two or more Epileptologists into patterns such as Normal/Slow, Highly Epileptiform Patterns (non-seizure) (HEP), Seizure, or Status Epilepticus (SE). The reviewer’s label that best represented a consensus among all reviewers was chosen as the final label for a file. The most severe category among all the epochs in the consensus label was assigned as the ground truth categorization of a file. The latest version of Clarity was run post-hoc on each of these recordings to obtain a maximum SzB per file. We identified cases in which the ground truth was SE and all cases in which Clarity would provide an SE alert (maximum SzB ≥90%) to determine the frequency of true positive, false negative, and false positive cases.

Results: In this dataset, there were 21 recordings with SE as the ground truth categorization and Clarity output indicated seizure burden ≥ 90% for 20 of them, demonstrating a sensitivity of 95%, specificity of 97%, negative predictive value of 99.9% and positive predictive value of 39% for SE detection. Of the cases that had seizure burden greater than 90% and were not categorized as SE, three-fourths of the cases were categorized as either seizures or highly epileptiform activity. Out of 28 recordings with ground truth categorization of seizures, not meeting criteria for SE, 3 were missed by the algorithm (Clarity seizure burden was 0) showing a sensitivity of 89.3% and specificity of 71% for seizure detection.

Conclusions: Clarity has a high level of sensitivity and specificity to identify suspected SE. It also has high sensitivity to short seizures and high negative predictive value to rule out SE or seizures.

These results suggest that Clarity can accurately monitor suspected SE in acute care practice, providing initial guidance to the bedside team to either escalate care or rule-out ongoing seizures.

Funding: Funded by Ceribell, Inc.