Automated Detection of Cardiac Arrhythmia Using Recurrent Neural Network

Irregular heartbeats, known as cardiac arrhythmia, are a medical problem. The World Health Organisation reports that cardiovascular disorders account for over 17 million deaths annually across the globe. Roughly 31% of all fatalities occur in this region. The American Heart Association reports that cardiovascular diseases are responsible for one out of every three deaths in the United States. Cardiovascular disease accounts for a greater proportion of fatalities than cancer and chronic obstructive pulmonary disease all together. Atrial fibrillation affects about 2% to 3% of the population in Europe and North America, according to a research done in 2014. Adults are considered to have bradycardia if their heart rate is less than 60 beats per minute, and tachycardia if it is more than 100 beats per minute. Premature contraction occurs when the heart rate is abnormally high. Fevers or fibrillation describe an irregular heartbeat. Cardiac arrhythmia can be classified using a variety of criteria, depending on factors other than heart rate alone. The location of the abnormal cardiac rhythm is another criterion for categorization. An erratic heartbeat is known as cardiac arrhythmia. In this study, we aim to demonstrate how to use deep learning algorithms to detect cardiac arrhythmia from electrocardiogram (ECG) signals with as little pre-processing as possible. Our automated anomaly detection system makes use of a 1D convolutional neural network (CNN) and long short-term memory (LSTM). Arrhythmia illness prediction using convolutional neural network (CNN) and long short-term memory (LSTM) algorithms is the primary goal of this work. The MIT-BH dataset, which includes seven distinct illness phases, is utilised to train both algorithms. Additionally, the accuracy was low with the current LSTM. In order to train and test for arrhythmia disease, this work used the CNN model.