Classification of Ventricular Arrhythmia using Machine Learning and Deep Learning Techniques

ECG is one of the most important and commonly used tools to detect any kind of abnormality in the heart. One such abnormality is arrhythmia, an irregular pattern which can lead to sudden cardiac arrest and even death. The number of deaths caused by cardiac arrest, which is a result of arrhythmia, has recently increased, and keeping this in view various models have been developed using machine learning and deep learning techniques and have worked on similar terms to provide better results in terms of accuracy. There are many forms of arrhythmias, and this work focuses on the methods used to classify one of the most dangerous forms of arrhythmias which is Ventricular Arrhythmia based on the ECG datasets collected from MIT-BIH Arrhythmia Database, MIT-BIH Supraventricular Arrhythmia Database, INCART 12-lead Database. The study presents insights into classification of ventricular arrhythmia using models such as Convolutional Neural Network (CNN), Dense Neural Network (DNN) and Long Short Term Memory (LSTM) and the way they are trained with beats and labels that are then used for classification. The study focuses on comparing the classification results of all the three techniques based on accuracy, AUC score, recall, prevalence, precision and specificity in order to conclude the best classifier among them in terms of various performance parameters so that it can be adapted for clinical use.