Robust detection of premature ventricular contractions using a wave-based Bayesian framework.

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Electrocardiogram (ECG) analysis is crucial for diagnosing heart conditions, especially in critical care and Holter monitoring.
• Accurate detection of premature ventricular contractions (PVCs) is vital for identifying life-threatening arrhythmias.

Purpose of the Study:

• To introduce a model-based dynamic algorithm for tracking ECG waveforms and detecting PVCs.
• To enhance the accuracy and reliability of cardiac rhythm monitoring.

Main Methods:

• Utilized an extended Kalman filter for tracking ECG characteristic waveforms.
• Introduced a "polargram" (polar signal representation) based on Bayesian estimations.
• Developed a novel signal fidelity measure using the covariance matrix of innovation signals.
• Detected PVCs by simultaneously tracking signal fidelity and the polar envelope.

Main Results:

• Achieved an average detection accuracy of 99.10% across diverse ECG databases.
• Demonstrated aggregate sensitivity of 98.77% and positive predictivity of 97.47% for PVC detection.
• Attained 100% accuracy for records containing only normal sinus beats and PVCs.

Conclusions:

• The proposed algorithm effectively detects and classifies ventricular complexes, including PVCs.
• The method enhances clinical PVC detection performance, contributing to improved patient care.
• The dynamic algorithm offers a robust approach for single or multi-lead ECG analysis.