Arrhythmia detection and classification using convolutional neural network
Healthcare and Life Sciences are among severely researched domains. With the introduction of computing paradigms and possibility of leveraging computational techniques have opened new research avenues. Heart is one among the most researched organs and the reason being trivial. With the advent of computational paradigms researchers explored ways to measure the electrical activity of the heartbeat which is known as electrocardiogram (ECG). Since then ECG has become one of the most important as well as primary tests to diagnose any irregularities in the functioning of the heart. The availability of the ECG data and possibility of employing deep learning models and their robustness has made researchers venture into leveraging them to elevate the accuracy of the ECG Analysis. To date there exists no end-to-end evaluation model based on deep learning techniques. In our present work, we identify different classes of cardiac rhythm by leveraging single-lead ECG. Our results show a significant improvement in ROC (approx. 0.97). Furthermore, our results demonstrate that we can classify a wide spectrum of arrhythmias using Deep Learning Techniques which are comparable to that of cardiologists. This approach can be used to minimize the component of human misdiagnose and help improve the quality of cardiac care.