A Convolutional Neural Network for Identifying Premature Ventricular Contraction Beat and Right Bundle Branch Block Beat

Classification of electrocardiograms (ECG) signals is obligatory for the automatic diagnosis of cardiovascular disease (CVD). Currently, there are a lot of machine learning methods which can be utilized for analyzing and classifying the ECG beats. However, the foremost shortcomings of these traditional machine learning solutions are use of a shallow structured classifier and hand-crafted suitable features. In this paper, we propose a nine-layer convolutional neural network (CNN) that can automatically extract appropriate features and detect the different categories of ECG beats based on individual records. Ten-fold cross-validation strategy is designed that boosts the robustness of our proposed CNN model. Our dataset consists of three types of ECG beats, namely: normal beat (N), premature ventricular contraction beat (V), and right bundle branch block beat (R). The experimental results demonstrate that with an appropriate choice of structure and parameters, our proposed deep learning model can classify the ECG hearts with a diagnose sensitivity of 98.37%, specificity of 99.19%, and accuracy of 98.92%. This proposed algorithm model can serve as an auxiliary tool to assist the clinicians for their diagnosis.