Ventricular fibrillation and tachycardia classification using a machine learning approach.

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Cardiovascular Diagnostics

Background:

• Accurate detection of ventricular fibrillation (VF) and rapid ventricular tachycardia (VT) is critical for automated external defibrillators and patient monitoring.
• Existing methods may have limitations in classification accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate a machine learning-based algorithm for classifying VF and VT using electrocardiogram (ECG) data.
• To optimize the algorithm's performance through feature selection and parameter tuning.

Main Methods:

• Utilized a support vector machine (SVM) for VF/VT classification.
• Extracted 14 metrics from ECG data within a specific window length.
• Employed a genetic algorithm for optimal variable selection.
• Trained and validated the algorithm on three public ECG databases, testing window sizes from 1 to 10 seconds.

Main Results:

• Achieved 98.1% accuracy, 98.4% sensitivity, and 98.0% specificity on in-sample training data using a 5-second window and two selected metrics.
• On out-of-sample validation data, obtained 96.3% accuracy, 96.2% sensitivity, and 96.2% specificity via fivefold cross-validation.
• The algorithm's performance surpassed current reported methods.

Conclusions:

• The proposed machine learning algorithm demonstrates high efficacy in classifying VF and VT.
• This method holds significant potential for improving the reliability of automatic external defibrillators and patient monitoring systems.
• The optimized feature selection and SVM approach offer a robust solution for cardiac arrhythmia detection.