Premium
Arrhythmia identification and classification using wavelet centered methodology in ECG signals
Author(s) -
Arumugam Maheswari,
Sangaiah Arun Kumar
Publication year - 2020
Publication title -
concurrency and computation: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.309
H-Index - 67
eISSN - 1532-0634
pISSN - 1532-0626
DOI - 10.1002/cpe.5553
Subject(s) - pattern recognition (psychology) , wavelet , feature extraction , noise (video) , artificial intelligence , signal (programming language) , computer science , energy (signal processing) , wavelet transform , identification (biology) , cardiac arrhythmia , feature (linguistics) , speech recognition , mathematics , statistics , medicine , cardiology , botany , philosophy , image (mathematics) , biology , programming language , atrial fibrillation , linguistics
Summary A systematic and profound reading of an electrocardiogram (ECG) is needed to identify the different kinds of cardiac diseases called Arrhythmia. The manual identification of the changes in the ECG pattern over a long period is challenging. This work can be automatized by developing algorithms that run perfectly on a computer or on a smartphone to identify the causes of arrhythmia. The proposed work includes three stages of analysis: (1) the ECG noise suppression, (2) RR and PR intervals extraction from the ECG signal, and the (3) ECG classification. The proposed methodology accurately identified the locations and amplitudes of P, Q, R, S, and T subwaves of the ECG signal using a dedicated wavelet design. Experimental results of the MIT‐BIH arrhythmia database records indicate the energy levels of the ECG signal at a decomposition level of 4 and 8 as 3.694 e +09 and 7.148 e +09 , respectively. These energy levels are used in deciding the wavelet decomposition levels for feature extraction and classification of the ECG signal. A decomposition level of eight is proposed in this work for perfect feature extraction and classification of the ECG signal. An analysis of subband frequencies obtained in the decomposition of the ECG signal is also performed. The proposed methodology gives a sensitivity of 99.58% and positive predictive value of 95.92% in the ECG examination.