Literature DB >> 26831270

Information theoretic multiscale truncated SVD for multilead electrocardiogram.

L N Sharma1.   

Abstract

BACKGROUND AND
OBJECTIVE: In this paper an information theory based multiscale singular value decomposition (SVD) is proposed for multilead electrocardiogram (ECG) signal processing. The shrinkage of singular values for different multivariate multiscale matrices at wavelet scales is based on information content. It aims to capture and preserve the information of clinically important local waves like P-waves, Q-waves, T-waves and QRS-complexes.
METHODS: The information is derived through clinically relevant multivariate multiscale entropy in SVD domain modifying Shannon's entropy. This optimizes the approximate ranks for matrices to capture the clinical components of ECG signals appearing at different scales. A newly introduced multivariate clinical distortion (MCD) metric is computed and compared with existing subjective and objective signal distortion measures. The proposed method is tested with records from CSE multilead measurement library and PTB diagnostic ECG database for various pathological cases.
RESULTS: It gives average percentage root mean square difference (PRD), average normalized root mean square error (NRMSE), average wavelet energy based diagnostic distortion measure (WEDD) values 5.8879%, 0.0059 and 1.0760% respectively for myocarditis pathology. The corresponding MCD value is 1.9429%. The highest average PRD and average WEDD values are 11.4053% and 5.5194% for cardiomyopathy with the corresponding MCD value 1.4003%.
CONCLUSIONS: Based on WEDD values and mean opinion scores (MOS), the quality group of all processed signals fall under excellent category.
Copyright © 2016 Elsevier Ireland Ltd. All rights reserved.

Entities:  

Keywords:  Multilead ECG; Multiscale SVD; Multivariate multiscale entropy; PRD; RMSE

Mesh:

Year:  2016        PMID: 26831270     DOI: 10.1016/j.cmpb.2016.01.010

Source DB:  PubMed          Journal:  Comput Methods Programs Biomed        ISSN: 0169-2607            Impact factor:   5.428


  1 in total

1.  GRMDA: Graph Regression for MiRNA-Disease Association Prediction.

Authors:  Xing Chen; Jing-Ru Yang; Na-Na Guan; Jian-Qiang Li
Journal:  Front Physiol       Date:  2018-02-20       Impact factor: 4.566

  1 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.