Limits...
Heart rate variability analysis using robust period detection.

Skotte JH, Kristiansen J - Biomed Eng Online (2014)

Bottom Line: Log transformed LF/HF values for series with distorted/removed beats were compared to undistorted values by linear regression.Similarly, the comparison between series with removed and undistorted beats yielded goodness of fit, coefficient and intercept of (0.98, 0.96, -0.01), (0.93, 0.78, -0.02) and (0.98, 0.95, 0.19) for RPD, FFT and LSP, respectively.The RPD method demonstrated superior performance compared to the FFT and LSP method by estimation of power spectral characteristics for HRV analysis.

View Article: PubMed Central - PubMed

Affiliation: National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100 Copenhagen, Denmark. jhs@nrcwe.dk.

ABSTRACT

Objective: Heart rate variability (HRV) analysis, which is an important tool for activity assessment of the cardiac autonomic nervous system, very often includes the estimation of power spectra for series of interbeat intervals (IBI). Ectopic beats and artifacts have a destructive effect on the standard methods (Fourier transform, FFT) for frequency analysis. This study investigates an alternative method for calculation of the periodogram using a robust period detection (RPD).

Method: Error free IBI series of 5 minutes for 221 subjects during one day were artificially distorted by randomly changing IBI values by ±15-40%. The low to high frequency rate (LF/HF) were calculated from periodograms estimated by the FFT, RPD and Lomb (LSP) methods for both error free and distorted series and for series with removed beats. Log transformed LF/HF values for series with distorted/removed beats were compared to undistorted values by linear regression.

Results: For series with 10% of distorted IBI values the regression analysis between distorted and undistorted series showed a goodness of fit, coefficient and intercept of 0.98, 0.94 and 0.02, respectively. In comparison, the values of these parameters were (0.34, 0.46, -1.61) and (0.28, 0.42,-1.32) for the FFT and LSP methods, respectively. Similarly, the comparison between series with removed and undistorted beats yielded goodness of fit, coefficient and intercept of (0.98, 0.96, -0.01), (0.93, 0.78, -0.02) and (0.98, 0.95, 0.19) for RPD, FFT and LSP, respectively.

Conclusion: The RPD method demonstrated superior performance compared to the FFT and LSP method by estimation of power spectral characteristics for HRV analysis.

Show MeSH

Related in: MedlinePlus

Examples of increasing amount of distortion applied to a 5 minutes series of IBI values. The series consists of 365 beats. The distortion rate increases from 0.5% in the uppermost graph to 15.4% in the bottom graph; red spikes represent distorted values.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4228159&req=5

Fig1: Examples of increasing amount of distortion applied to a 5 minutes series of IBI values. The series consists of 365 beats. The distortion rate increases from 0.5% in the uppermost graph to 15.4% in the bottom graph; red spikes represent distorted values.

Mentions: The data set consisted of all 5 minutes periods from 221 subjects monitored for HR during 1 day. Approximately 50% of the 5 minutes periods were without abnormal beats according to the criterion that normal IBI intervals should not deviate more than 15% from its neighboring intervals. Thirty percent of the 5 minutes periods were found to include errors but with an amount less than 10%. Figure 1 shows an example of a 5 minutes period without abnormal beats that has been processed by the ectopic distortion algorithm to induce random errors of varying ectopic strength.Figure 2 shows a plot of the LF/HF ratio of power spectra estimated from IBI series where approximately 1% of the IBIs have been changed corresponding to a displacement of the beats with a distance of ±15-40% to its neighbors. The LF/HF ratio for the ectopic distorted IBI series are compared with the undistorted series calculated by the RPD, FFT and LSP methods. It is evident that the small amount of distorted beats (typically 1–3 beats in an 5 minutes period, 1% of distorted IBIs corresponds to 0.5% distorted beats) have a large influence on the FFT and LSP calculated periodogram and very little influence on the RPD results.Figure 1


Heart rate variability analysis using robust period detection.

Skotte JH, Kristiansen J - Biomed Eng Online (2014)

Examples of increasing amount of distortion applied to a 5 minutes series of IBI values. The series consists of 365 beats. The distortion rate increases from 0.5% in the uppermost graph to 15.4% in the bottom graph; red spikes represent distorted values.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4228159&req=5

Fig1: Examples of increasing amount of distortion applied to a 5 minutes series of IBI values. The series consists of 365 beats. The distortion rate increases from 0.5% in the uppermost graph to 15.4% in the bottom graph; red spikes represent distorted values.
Mentions: The data set consisted of all 5 minutes periods from 221 subjects monitored for HR during 1 day. Approximately 50% of the 5 minutes periods were without abnormal beats according to the criterion that normal IBI intervals should not deviate more than 15% from its neighboring intervals. Thirty percent of the 5 minutes periods were found to include errors but with an amount less than 10%. Figure 1 shows an example of a 5 minutes period without abnormal beats that has been processed by the ectopic distortion algorithm to induce random errors of varying ectopic strength.Figure 2 shows a plot of the LF/HF ratio of power spectra estimated from IBI series where approximately 1% of the IBIs have been changed corresponding to a displacement of the beats with a distance of ±15-40% to its neighbors. The LF/HF ratio for the ectopic distorted IBI series are compared with the undistorted series calculated by the RPD, FFT and LSP methods. It is evident that the small amount of distorted beats (typically 1–3 beats in an 5 minutes period, 1% of distorted IBIs corresponds to 0.5% distorted beats) have a large influence on the FFT and LSP calculated periodogram and very little influence on the RPD results.Figure 1

Bottom Line: Log transformed LF/HF values for series with distorted/removed beats were compared to undistorted values by linear regression.Similarly, the comparison between series with removed and undistorted beats yielded goodness of fit, coefficient and intercept of (0.98, 0.96, -0.01), (0.93, 0.78, -0.02) and (0.98, 0.95, 0.19) for RPD, FFT and LSP, respectively.The RPD method demonstrated superior performance compared to the FFT and LSP method by estimation of power spectral characteristics for HRV analysis.

View Article: PubMed Central - PubMed

Affiliation: National Research Centre for the Working Environment, Lersø Parkallé 105, DK-2100 Copenhagen, Denmark. jhs@nrcwe.dk.

ABSTRACT

Objective: Heart rate variability (HRV) analysis, which is an important tool for activity assessment of the cardiac autonomic nervous system, very often includes the estimation of power spectra for series of interbeat intervals (IBI). Ectopic beats and artifacts have a destructive effect on the standard methods (Fourier transform, FFT) for frequency analysis. This study investigates an alternative method for calculation of the periodogram using a robust period detection (RPD).

Method: Error free IBI series of 5 minutes for 221 subjects during one day were artificially distorted by randomly changing IBI values by ±15-40%. The low to high frequency rate (LF/HF) were calculated from periodograms estimated by the FFT, RPD and Lomb (LSP) methods for both error free and distorted series and for series with removed beats. Log transformed LF/HF values for series with distorted/removed beats were compared to undistorted values by linear regression.

Results: For series with 10% of distorted IBI values the regression analysis between distorted and undistorted series showed a goodness of fit, coefficient and intercept of 0.98, 0.94 and 0.02, respectively. In comparison, the values of these parameters were (0.34, 0.46, -1.61) and (0.28, 0.42,-1.32) for the FFT and LSP methods, respectively. Similarly, the comparison between series with removed and undistorted beats yielded goodness of fit, coefficient and intercept of (0.98, 0.96, -0.01), (0.93, 0.78, -0.02) and (0.98, 0.95, 0.19) for RPD, FFT and LSP, respectively.

Conclusion: The RPD method demonstrated superior performance compared to the FFT and LSP method by estimation of power spectral characteristics for HRV analysis.

Show MeSH
Related in: MedlinePlus