Open AccessAutomated Coronary Artery Disease Detection using RQA Features and Quadratic Support Vector Machine
Author(s) -
Suvra Mandal,
Nabanita Sinha
Publication year - 2020
Publication title -
international journal of recent technology and engineering
Language(s) - English
Resource type - Journals
ISSN - 2277-3878
DOI - 10.35940/ijrte.e6221.018520
Subject(s) - cad , support vector machine , recurrence plot , recurrence quantification analysis , artificial intelligence , coronary artery disease , pattern recognition (psychology) , heart rate variability , quadratic classifier , cardiology , classifier (uml) , computer science , medicine , nonlinear system , heart rate , engineering , physics , quantum mechanics , engineering drawing , blood pressure
One of the major causes of death globally due to heart disease is the coronary artery disease (CAD). Due to CAD the blood flow to the cardiac muscle is reduced. The progression of this process eventually causes Myocardial Infarction (MI) that result in sudden death. Hence the detection of CAD at early phase is essential. The electrocardiogram (ECG) is mainly used to capture the abnormal cardiac activity for CAD. But the difficulties in manual interpretation of ECG signal leads to error in CAD detection. To overcome the difficulty in CAD diagnostic task, we have proposed a computer aided methodology using the heart rate variability signal (HRV) for auto diagnosis of CAD and Normal heart condition. The hidden characteristics of HRV signal are identified through Recurrence Plot (RP) and the hidden information is quantified by Recurrence quantification analysis (RQA). The extracted RQA based nonlinear features are analyzed for their clinical significance. The set the effective features are used for classification and subjected to three types of Support vector machine (SVM) classifier to discriminate CAD and the normal heart condition. The ECG database of CAD and normal subjects are taken from Physio.net database to obtain the experimental results. The highest diagnostic ability of the classifier is obtained by quadratic SVM with the accuracy of 98.83% where as the linear and cubic SVM classifier provide 97.22 % and 98.37 % classification accuracy respectively.