Clinical Data and Supporting Evidence: ClarityTM Study
CERIBELL WITH CLARITY WAS DEMONSTRATED TO DIFFERENTIATE EEGS WITH AND WITHOUT STATUS EPILEPTICUS WITH HIGH SENSITIVITY AND SPECIFICITY.75
Understanding the Clarity AI Algorithm
Clarity is a machine learning algorithm within the Ceribell system that has been trained on thousands of hours of expert annotated critical care EEGs with the goal of alerting to or ruling out life-threatening continuous seizures also known as status epilepticus.
Study Design
NEW DATA from a larger retrospective assessment of 665 adult patients who underwent evaluation of possible seizures with Point-of-Care EEG system
Expert consensus
Majority agreement EEG reading from multiple EEG experts
The study assessed the performance of Ceribell
with Clarity to triage seizure activity in critical care
and emergency department settings
Clarity
Ceribell with Clarity EEG interpretation
Current standard of care
Individual epileptologist’s EEG reading
Key Results
Clarity for detection of status epilepticus
Clarity detected ≥ 90% seizure
burden, thereby triggering an alert for impending status epilepticus, in 19 out of 20 cases (95% sensitivity)
Clarity for ruling out seizure
Of the 450 EEG recordings in which
Clarity detected no seizures, expert reviewers identified seizures in only 4 cases.
Seizure Burden Yield
- 5% of EEGs resulted in a bedside alert (had a seizure burden ≥ 90%)
- 27% of EEGs had seizure or epileptiform abnormalities that did not result in a bedside alert (seizure burden between 1%-90%)
- 68% of EEGs were identified as normal or slow (0% seizure burden)
Inter-rater Variability
All 4 neurologists have fellowship training in epilepsy
Human neurologists show variability in detection of status epilepticus
