Automated Annotation of Epileptiform Burden and Its Association with Outcomes

Area of Science:

• Neuroscience
• Clinical Neurology
• Medical Informatics

Background:

• Epileptiform activity is a common finding in acutely ill patients, but its precise impact on outcomes is not fully understood.
• Quantifying epileptiform activity burden has been challenging due to the complexity of electroencephalography (EEG) data.

Purpose of the Study:

• To determine the dose-response relationship between epileptiform activity burden and clinical outcomes in acutely ill patients.
• To establish peak epileptiform activity burden as a quantifiable predictor of patient outcomes.

Main Methods:

• Retrospective analysis of 1,967 patients with >16 hours of continuous EEG monitoring.
• Development of an artificial intelligence algorithm to quantify epileptiform activity burden from large EEG datasets.
• Machine learning models used to assess the association between epileptiform burden and modified Rankin Scale (mRS) at discharge.

Main Results:

• Peak epileptiform activity burden was independently associated with poor outcomes (mRS 5-6) (p < 0.0001).
• An increase in peak epileptiform activity burden from 0% to 100% increased the probability of a poor outcome by 35%.
• Other independent predictors of poor outcome included age, APACHE II score, seizure on presentation, and hypoxic-ischemic encephalopathy.

Conclusions:

• Automated measurement of peak epileptiform activity burden provides a consistent and quantifiable metric.
• This metric serves as a valuable target for future clinical trials investigating interventions to suppress epileptiform activity.
• Findings support the clinical significance of quantifying epileptiform activity burden for predicting patient outcomes.