Advanced ECG feature extraction and SVM classification for predicting defibrillation success in OHCA

Area of Science:

• Biomedical Engineering
• Cardiology
• Emergency Medicine

Background:

• Out-of-hospital cardiac arrest (OHCA) poses a significant challenge to emergency medical services.
• Rapid and accurate prediction of defibrillation outcomes is essential for improving patient survival rates.

Purpose of the Study:

• To develop a non-invasive, rapid, and reliable method for predicting defibrillation success in OHCA patients.
• To leverage electrocardiogram (ECG) signal features and machine learning for enhanced clinical decision-making during resuscitation.

Main Methods:

• Extracted six ECG features (heart rate, QRS amplitude/duration, total/low/high-frequency power) from 251 OHCA patient signals.
• Identified QRS complex amplitude, total power, and low-frequency power as the most discriminative features using AUC values.
• Trained a Support Vector Machine (SVM) classifier on selected features to predict defibrillation success.

Main Results:

• The SVM classifier achieved a prediction accuracy of 95.6% for defibrillation success.
• QRS complex amplitude, total power, and low-frequency power demonstrated high discriminative ability.
• The study successfully combined targeted ECG features with machine learning for accurate outcome prediction.

Conclusions:

• ECG-based feature analysis coupled with machine learning offers a promising approach to forecast defibrillation success in OHCA.
• This method can support clinical decisions, potentially increasing survival rates for OHCA patients.
• Future research will focus on expanding datasets and exploring advanced machine learning models to further improve prediction accuracy.