Predicting the Higher Energy Need for Effective Defibrillation Using Machine Learning Based on an Animal Model

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Early defibrillation is crucial for cardiac arrest survival, but optimal energy levels are debated.
• Arterial blood gas (ABG) parameters are often available in clinical settings.
• This study explored ABG's predictive power for defibrillation energy requirements.

Purpose of the Study:

• To investigate if ABG parameters can predict defibrillation energy needs.
• To identify ABG markers associated with defibrillation threshold (DFT).
• To develop a machine learning model for personalized defibrillation energy selection.

Main Methods:

• Ventricular fibrillation induced in an animal model to determine DFT.
• ABG parameters (PaCO2, PaO2, pH, Hct, Na+, K+, HCO3-) measured pre-defibrillation.
• Classical analysis and machine learning models used to correlate ABG with DFT.

Main Results:

• Hematocrit (Hct) and sodium (Na+) levels differed significantly between high (>130 J) and low (<40 J) DFT categories.
• DFT negatively correlated with PaO2 and positively with Hct and Na+.
• Extra Trees Classifier model achieved 83% accuracy, predicting higher energy needs, with Hct, PaCO2, and PaO2 as key predictors.

Conclusions:

• ABG parameters, analyzed with modern data techniques, can guide personalized defibrillation energy selection.
• This approach is particularly relevant for controlled clinical environments like ICUs and cath labs.
• Predictive modeling enhances first-shock success rates in cardiac arrest management.