Enhancing Arrhythmia Classification and Prediction using Deep Learning

Arrhythmia is a condition that develops when the heart's electromagnetic impulses do not function properly. The heart has an irregular rhythm during an arrhythmia. An electrocardiogram (ECG) is a typical instrument for evaluating heart activity and diagnosing arrhythmias. Arrhythmia is related to sudden cardiac death and the disease is a leading cause of death in the universe. It is mainly affected in the young and middle adult people. This paper explores and compares multiple explanations to the problem of the effectiveness of data augmentation in arrhythmia ECG signal sensor classification. The efficiency of data augmentation using basic methods like jittering scaling and time-warping input signal has been shown. The paper assesses various data augmentation techniques by artificially restricting access to a section of the publicly available MIT-BIH dataset. The above-mentioned are the data augmentation techniques. The proposed model that we are employed is built on a combination of CNN, long short-term memory (LSTM) + convolutional neural networks (CNN), and CNN+LSTM+Attention model that are trained using multiple classifiers and a mean square loss function. By using these 3 techniques the results show that accuracy of CNN model is 99.16, CNN+LSTM is 99.33, CNN+LSTM+Attention model is 99.36%.