Characterization of coronary artery disease using flexible analytic wavelet transform applied on ECG signals. (January 2017)

• Record Type: Journal Article
• Authors: Kumar, Mohit; Pachori, Ram Bilas; Acharya, U. Rajendra
• Is Part Of: Biomedical signal processing and control. Volume 31(2017)
• Page Start: 301
• Page End: 308
• Publication Date: 2017-01
• Subjects: Coronary artery disease -- Flexible analytic wavelet transform -- ECG beats -- Cross information potential -- Student's t-test; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2016.08.018
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract : Highlights: Classification of CAD and normal ECG signals is proposed. ECG signals are segmented into beats. FAWT is used to decompose the ECG beats. CIP is computed from the detail coefficients of FAWT decomposition. Obtained classification accuracy is 99.60%. Abstract: In the present work, an automated diagnosis of Coronary Artery Disease (CAD) using Electrocardiogram (ECG) signals is proposed. First, the ECG signals of 40 normal subjects and 7 CAD subjects are segmented into beats. 137, 587 ECG beats of normal subjects and 44, 426 ECG beats of CAD subjects are used in this work. Flexible Analytic Wavelet Transform (FAWT) technique is used to decompose the ECG beats. Cross Information Potential (CIP) parameter is computed from the real values of detail coefficients of FAWT based decomposition. For CAD subjects mean value of CIP parameter is found higher in comparison to normal subjects. Thereafter, Student's t -test method and Kruskal–Wallis statistical test are applied to check the discrimination ability of the extracted features. Further, the features are fed to Least Squares-Support Vector Machine (LS-SVM) for performing the classification. Classification accuracy is computed at every decomposition level starting from the first level of decomposition. We have observed significant improvement in the classification accuracy up to fourth level of decomposition. At fifth level of decomposition classification accuracy is not improved significantly as compared to the fourth level of decomposition. Hence, we analysed the ECG beats up to fifth level of decomposition. Accuracy of classification is higher for Morlet wavelet kernel (99.60%) in comparison to Radial Basis Function (RBF) kernel (99.56%). The developed methodology can be used in mass cardiac screening and can aid cardiologists in performing diagnosis.