"Early Stratification of Risk for Poor Neurological Outcome After Cardiac Arrest Is Improved with Processed EEG Data"

Area of Science:

• Neurology
• Machine Learning
• Critical Care Medicine

Background:

• Cardiac arrest survivors often experience neurological deficits.
• Early and accurate neurological risk stratification is crucial for patient management.
• Quantitative EEG (qEEG) offers potential biomarkers for neurological assessment.

Purpose of the Study:

• To assess the impact of integrating early processed qEEG biomarkers with clinical data for neurological risk stratification post-cardiac arrest.
• To develop and compare machine learning models for predicting patient outcomes using different data combinations.

Main Methods:

• Collected clinical data and processed EEG metrics (suppression ratio, BIS) from comatose cardiac arrest patients.
• Developed six machine learning models to predict discharge and long-term outcomes.
• Compared model performance using clinical data alone, EEG data alone, and combined data.

Main Results:

• The combination of early processed EEG and clinical data yielded the highest predictive accuracy (AUC 0.88 for long-term outcome).
• Integrating EEG metrics significantly improved model performance compared to using clinical or EEG data independently (p<0.001).
• Machine learning models effectively stratified neurological risk in this patient cohort.

Conclusions:

• Early processed EEG biomarkers combined with clinical data provide superior neurological risk stratification after cardiac arrest.
• Machine learning algorithms effectively leverage combined data for outcome prediction.
• External validation is recommended to confirm these findings.