Wavelet analysis of heart rate variability: Impact of wavelet selection. (February 2018)

• Record Type: Journal Article
• Authors: Tzabazis, Alexander; Eisenried, Andreas; Yeomans, David C.; Hyatt, Moore IV
• Is Part Of: Biomedical signal processing and control. Volume 40(2018)
• Page Start: 220
• Page End: 225
• Publication Date: 2018-02
• Subjects: Electrocardiography -- Heart rate variability -- Autonomic nervous system -- Wavelet analysis -- Haar -- Daubechies -- Low frequency -- High frequency; 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.2017.09.027
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Heart rate variability analysis was performed using maximum overlap discrete wavelet packet transform. Correlations of results obtained with different kernels were calculated. Some correlations were clinically insignificant, especially for the LF/HF ratio. Kernels used for wavelet analysis need to be reported to make results comparable. Specific kernels might perform better depending on application. Abstract: Background: Wavelet transform based analysis of heart rate variability is increasingly being used for a wide variety of clinical applications. There is no gold standard as to which wavelet to use and the correlation between results obtained by using different wavelets is unknown. Methods: Heart rate variability in electrocardiograms from healthy volunteers was analyzed using the following wavelets (maximum overlap discrete wavelet packet transform): Haar, Daubechies 2, 4, and 8, least asymmetric Daubechies 4 and 8, and best localized Daubechies 7 using the RHRV package in R. Correlation of power in the different frequency bands (ultra low frequency (ULF), very low frequency (VLF), low frequency (LF), high frequency (HF)) as well as total power and LF:HF ratio were calculated. Bland-Altman comparisons were also made for selected wavelets to test for agreement. Findings: Correlations between results obtained by different wavelets were all statistically significant. Most correlation coefficients were moderate (0.3 ≤ r ≤ 0.7). They were, however, generally lower for the LF:HF ratio, which is commonly used to assess balance of the autonomic nervous system. Conclusion: It is necessary to report which wavelet is used when performing wavelet transform based heart rate variability analysis and depending on whether one is interested in detecting onset or intensity of changes performance of wavelets will vary.