Automated diagnosis of premature ventricular contraction arrhythmia through electrocardiogram analysis and machine learning techniques

Premature ventricular beat and its attacks are one of the most common abnormalities that not only old people but also about 33% of heart diseases suffer from. Many studies have proven that cardiovascular diseases not only have high morbidity and mortality, but also have significant psychological impacts, which lead to a high disease burden of these diseases. Therefore, early diagnosis and treatment of these diseases can greatly reduce the disease burden. In this study, we presented an automated framework based on morphological and wavelet-based features extracted from electrocardiogram (ECG) signals and a hybrid classification approach based on three classifiers: multilayer perceptron (MLP), Support Vector Machine (SVM), and K-nearest neighbors (KNN). Our framework was tested on the MIT-BIH database and achieved 99.53% accuracy, 100% sensitivity, and 99.12% specificity for PVCA detection. Also, our proposed approach achieved acceptable results in detecting PVCA heartbeats from single ECG recordings. Therefore, this system can be considered as a candidate for clinical systems to diagnose PVCA.