Diagnosis of Coronary Artery Disease using 1-D Convolutional Neural Network

Abstract

Heart disease is treated as one of the noxious diseases at the present time. Coronary artery disease is a kind of heart syndrome which is statistically growing day by day in the society. It is very tough for medical practitioners to predict Coronary artery disease as it is a complicated task that needs experience and acquaintance. For the detection of the disease doctors normally prescribe various invasive and non-invasive methods like angiography, ECG and echocardiogram. These methods are very expensive and sometimes not able to discover a number of undiagnosed symptoms. Due to these, it is not possible to detect the disease accurately at an early stage. The medical sector today contains a number of useful data that is helpful to detect a disease accurately. Using this data many researchers have proposed a number of intelligent systems for the detection of the disease. In this work, a competent system is implemented using deep learning for the better detection of the disease. The system is constructed using Convolutional Neural Network. The system has three phases. In the first phase data cleaning, data imputation and important feature selection are performed. In the second phase model training and hyperparameter tuning is performed. Finally, in the last phase, the model prediction is performed using the test data. The data set used for experimentation is Cleveland, Hungary, Switzerland and Long beach heart disease data present in the UCI repository. The proposed system gives a classification accuracy of 96.49% during testing, which is highest among all the discussed methods.