Machine Learning for Classification and Control of Cardiac Arrhythmias

Abstract

Cardiac arrest has become a primary cause of sudden death. Commonly it is caused by cardiac arrhythmias such as Tachycardia and Bradycardia. The aim of the paper is to propose machine learning algorithm for classifying the arrhythmias accurately using Electrocardiograph (ECG) signals and Clinical data. Further, a suitable model-based control scheme is developed for controlling Bradycardia. Firstly, the best features are extracted from the data set and are used for classifications of cardiac arrhythmias using Convolution Neural Network (CNN), Support Vector Machine (SVM) and CNN-SVM (SVM). The classification accuracy is compared for the proposed Machine Learning Algorithms with training dataset and test dataset. Secondly, after classifying cardiac arrhythmia, suitable model is identified by developing various nonlinear models namely Nonlinear Auto Regressive Exogenous (NARX), Nonlinear Hammerstein-Wiener (HW) and Recurrent Neural Network (RNN) for cardiac vascular system. The performances of the developed models are compared, and best model is intended to design controller. Finally, model-based control scheme is developed using the best model and closed loop studies are carried out. The simulation studies show the feasibility of the proposed control scheme.