ECG signal analysis using CWT, spectrogram and autoregressive technique

• Published in: Iran Journal of Computer Science (Online) Vol. 4; no. 4; pp. 265 - 280
• Main Authors: Gupta, Varun, Mittal, Monika, Mittal, Vikas, Gupta, Anshu
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2021
• Subjects: Artificial Intelligence; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Mathematics of Computing; Original Article; Software Engineering/Programming and Operating Systems; Theory of Computation; Preprocessing; Cardiovascular; Electrocardiogram; Continuous wavelet transform; Feature extraction; Nonlinear behavior; Autoregressive technique; Computer-aided diagnosis; Spectrogram
• ISSN: 2520-8438; 2520-8446
• DOI: 10.1007/s42044-021-00080-8

The cardiovascular system is a combination of the heart, blood and blood vessels. Cardiovascular diseases (CVD) are a key factor behind casualties worldwide among both women and men. About 9.4 million deaths occur due to high Blood Pressure (BP) only, out of which 51% deaths are due to strokes and 45% deaths are due to coronary heart diseases. The Electrocardiogram (ECG) represents the heart health condition of the subject, (patient) since it is acquired through electrical conduction, which appears in terms of P-QRS-T waves. But analysis of these waves is very tedious due to the existence of different noises/artifacts. Computer Aided Diagnosis (CAD) system is required in practical medical scenario for better and automated ECG signal analysis and to compensate for human errors. In general, implementation of a CAD system for ECG signal analysis requires; preprocessing, feature extraction and classification. In the existing literature, some authors have used time domain techniques which yield good performance for cleaned ECG signals i.e., without noise/artifact. Some authors have used frequency domain techniques later, but they suffer from the problem of spectral leakage making them unsuitable for real time/pathological datasets. The existing techniques from both these domains are not able to effectively analyze nonlinear behavior of ECG signals. These limitations have motivated this work where Continuous Wavelet Transform (CWT), Spectrogram and Autoregressive (AR) technique are used collectively for interpreting nonlinear and non-stationary features of the ECG signals. In this paper, both Massachusetts Institute of Technology-Beth Israel Hospital Arrhythmia database (MB Ar DB) and Real-time database (RT DB) have been used. Performance of the proposed method is compared with that of the previous studies on the basis of sensitivity (SE) and detection rate (D.R). The proposed technique yields SE of 99.90%, D.R of 99.81% & SE of 99.77%, D.R of 99.87% for MB Ar DB and RT DB, respectively. Therefore, the proposed technique showcases the possibility of an encouraging diagnostic tool for further improving the present situation of health informatics in cardiology labs/hospitals.