AUTOMATIC CLASSIFICATION OF ECG USING DEEP LEARNING MODEL

Abstract

Early detection of cardiovascular diseases through ECG signals has played an important role in the treatment process for patients. The application of artificial intelligence to develop an automatic method of classifying ECG signals with high accuracy and reliability to help reduce the time of diagnosis is an inevitable requirement. This study proposes a deep learning model that combines a Densely Connected Convolutional Networks (DenseNet) with a Bidirectional Long Short-Term Memory Networks (BiLSTM) with a small number of parameters into the ECG signal classification. The proposed model is evaluated on an open database consisting of 827 ECG records. Although the study only uses a small amount of data for the training process, the proposed model still gives good results, in particularly, F1 scores corresponding to the types of right bundle branch block, left bundle branch block and sinus bradycardia are 0.831, 0.846 and 0.882 respectively. The obtained results can serve as a basis for further studies applying on devices with limited resources.