Elsevier

Medical Engineering & Physics

Volume 35, Issue 12, December 2013, Pages 1778-1785
Medical Engineering & Physics

Application of an automatic adaptive filter for Heart Rate Variability analysis

https://doi.org/10.1016/j.medengphy.2013.07.009Get rights and content

Abstract

The presence of artifacts and noise effects in temporal series can seriously hinder the analysis of Heart Rate Variability (HRV). The tachograms should be carefully edited to avoid erroneous interpretations. The physician should carefully analyze the tachogram in order to detect points that might be associated with unlikely biophysical behavior and manually eliminate them from the data series. However, this is a time-consuming procedure. To facilitate the pre-analysis of the tachogram, this study uses a method of data filtering based on an adaptive filter which is quickly able to analyze a large amount of data. The method was applied to 229 time series from a database of patients with different clinical conditions: premature newborns, full-term newborns, healthy young adults, adults submitted to a very-low-calorie diet, and adults under preoperative evaluation for coronary artery bypass grafting. This proposed method is compared to the demanding conventional method, wherein the corrections of occasional ectopic beats and artifacts are usually manually executed by a specialist. To confirm the reliability of the results obtained, correlation coefficients were calculated, using both automatic and manual methods of ltering for each HRV index selected. A high correlation between the results was found, with highly significant p values, for all cases, except for some parameters analyzed in the premature newborns group, an issue that is thoroughly discussed. The authors concluded that the proposed adaptive filtering method helps to efficiently handle the task of editing temporal series for HRV analysis.

Introduction

The successive values in the instantaneous heart rate referred to as RR intervals or NN intervals [1], which are used in the Heart Rate Variability (HRV) analysis [2], [3], [4], are displayed in a plot (tachogram). The horizontal axis of this tachogram shows the events over time while the vertical axis displays the values of the intervals between heartbeats. In general, the analysis of tachograms or experimental time series requires that these series be properly filtered in a way that excludes ectopic beats or noise effects.

For sequential analysis of intervals between heartbeats it is essential to remove artifacts, impulses not initiated by the sinoatrial node (ventricular premature complexes – VPC) [5] and spurious interference such as those due to muscle tremors, poor placement of electrodes, atypical electronics effects on the environment to capture the signal, and so forth.

Fig. 1 shows two tachograms each with 1750 RR intervals. Fig. 1A contains a lot of artifacts that must be excluded (RR intervals with extremely high or low values), while Fig. 1B shows a variability compatible with clinical behavior.

In analysing HRV it is necessary that the heart rhythm be sinusal (the origin is from the sinus node). Atrial or ventricular extrasystoles (whereby heartbeats are initiated by the atrial or ventricular cells rather than by the sinus node) are not the deciding factor when it comes to the issue of excluding artifacts, but instead several other situations that can render a tachogram unsuitable for analysis. In these situations the “human eye” is considered the gold standard [6] to discern whether to use a particular RR tachogram pattern or not.

As discussed in previous articles Storck et al. [7] and Karlsson et al. [8], the use of a computational implemented preprocessing procedure could be a very convenient tool to aid an expert in the manual analyses, allowing one to accelerate the procedure of analyzing a large amount of data.

This adaptive filter can self-adjust its parameters according to the input signal and improve its analytical performance [9]. The main objective of this study is to apply and compare the adaptive filter method for time series of HRV analysis with the conventional filter method by considering different clinical situations. Such clinical situations include: premature newborns hospitalized in intensive care units, full-term newborns, healthy young adults, adults submitted to a very-low-calorie diet for the purpose of weight loss, and adults in the preoperative evaluation for coronary artery bypass grafting for severe coronary disease.

The comparison between the filtered series using conventional and adaptive methods is performed by statistical analysis of HRV indexes calculated in the frequency domain using nonlinear techniques (Poincaré plot and entropy-based analysis).

Section snippets

Material and methods

This section describes the HRV data used, the conventional and adaptive filters, and also statistical analysis of HRV features using frequency-domain and nonlinear descriptors (Poincaré plot, approximate entropy and sample entropy).

Results

Free parameters of the filter were chosen on an empirical basis but in accordance with and following the analyses carried out by a specialist, as a training process to setting the filter final parameters. The values of the parameters used are: c = 0.05, ρ = 10, a = 3 e σb = 0.02.

Two representative examples of application of the adaptive fiter are given below for two time series. The first time series is an example with relatively few artifacts and the second one is an example with a large number of

Discussion

In our analysis, the filter parameters have been adjusted to be used for the whole set of data (see supplementary data). This approach makes the filter more robust so that it could be used for all groups of RR time series (without ectopic beats), in addition to standardizing the use of the adaptive filter for all time series presented in this study.

Through comparison with the conventional method, which requires the intervention of an expert to manually remove ectopic beats in the processing of

Funding

None.

Conflict of interest statement

None declared.

Ethical approval

Not required.

Acknowledgments

The authors thank CAPES/Brazil and CNPq/Brazil for financial support.

References (43)

  • N. Storck et al.

    Automatic computerized analysis of heart rate variability filtering of ectopic beats

    Clinical Physiology

    (2001)
  • M. Karlsson et al.

    Automatic filtering of outliers in RR intervals before analysis of heart rate variability in holter recordings: a comparison with carefully edited data

    BioMedical Engineering OnLine

    (2012)
  • V.K. Ingle et al.

    Digital signal processing using Matlab

    (1997)
  • J. Bernardes et al.

    The porto system for automated cardiotocographic signal analysis

    Journal of Perinatal Medicine

    (1991)
  • H. Goncalves et al.

    Internal versus external intrapartum foetal heart rate monitoring: the effect on linear and nonlinear parameters

    Physiological Measurement

    (2006)
  • D.A. de Campos et al.

    Sisporto multicentre validation study group. prediction of neonatal state by computer analysis of fetal heart rate tracings: the antepartum arm of the sisporto multicentre validation study

    European Journal of Obstetrics & Gynecology and Reproductive

    (2005)
  • H. Goncalves et al.

    Linear and nonlinear fetal heart rate analysis of normal and acidemic fetuses in the minutes preceding delivery

    Medical & Biological Engineering & Computing

    (2006)
  • A. Costa et al.

    Prediction of neonatal acidemia by computer analysis of fetal heart rate and ST event signals

    American Journal of Obstetrics & Gynecology

    (2009)
  • F.A. Selig et al.

    Heart rate variability in preterm and term neonates

    Arquivos Brasileiros de Cardiologia

    (2011)
  • J.C. Leal et al.

    Perioperative serum troponin 1 levels are associated with higher risk for atrial fibrillation in patients undergoing coronary artery bypass graft surgery

    Interactive CardioVascular and Thoracic Surgery

    (2012)
  • F.X. Gamelin et al.

    Validity of the polar s810 heart rate monitor to measure RR intervals at rest

    Medicine & Science in Sports & Exercise

    (2006)
  • Cited by (0)

    View full text