Detection of Arrhythmia Using Convolutional Neural Networks

Abstract

AbstractCardiac arrhythmias are irregular heartbeats that are either too fast (tachycardia) or too slow (bradycardia). A minor alteration in the morphology or dynamics of the electrocardiogram (ECG) can induce severe arrhythmia events, which can impair the heart’s ability to pump blood and cause shortness of breath, chest pain, exhaustion, and loss of consciousness. A few types of arrhythmias are present infrequently in the ECG signal and, as a result, must be recognised using extensive ECG recordings. Longer ECG records necessitate manual analysis, which takes time and effort. The ANSI/AAMI EC57 guideline identified five categories in heartbeats: non-ectopic or paced beat (N), supraventricular ectopic beat (S), ventricular ectopic beat (V), fusion of a ventricular and normal beat (F), pace beat or fusion of a paced and normal beat (F), and beat which is not categorised (Q). To classify these five types of heartbeats, a technique based on deep neural network has proposed in the study. Physicians benefit greatly from the automatic detection and classification of cardiac arrhythmias. The proposed technique was tested with a patient-specific pattern on the MIT-BIH Arrhythmia Dataset. The overall accuracy for the proposed model achieved was 97.62% with F1-score of 0.88 which is a decent score. The proposed approach performed well in the experiments, indicating that it could be useful in clinical practice.KeywordsCardiac arrhythmiasElectrocardiogramConvolutional neural networks