Detection of Premature Ventricular Contractions Using Machine Learning

Premature Ventricular Contractions (PVCs), a form of abnormal heartbeat that can be identified through electrocardiogram (ECG) signals, play a crucial role in detecting potentially life-threatening ventricular arrhythmias. In this study, three features (RR interval, QRS width, and R amplitude) are extracted from the MIT-BIH Arrhythmia Database and used Decision Tree (DT), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) as classifiers. The classifiers achieved satisfactory results, with average accuracy rates of 94 % for KNN(K = 5) and 93% for KNN (K = 7), 87% for SVM, and 93% for DT. In addition, the classifiers were tested with the St. Petersburg Institute of Cardiological Technics 12-lead Arrhythmia database and obtained a convincing result of 74% accuracy for the SVM classifier, 70% for the KNN (K=5) and 68% KNN(K = 7) classifier, and 95% for the DT classifier. These results highlight the potential of feature selection and classification techniques in accurately identifying PVC beats from ECG signals, which is crucial for the early detection and effective treatment of ventricular arrhythmias.