BACKGROUND Premature beats (PBs) have been considered as artifacts producing a bias in the traditional analysis of heart rate (HR) variability. We assessed the effects and significance of PBs on fractal scaling exponents in healthy subjects and patients with a recent myocardial infarction (AMI). METHODS Artificial PBs were first generated into a time series of pure sinus beats in 20 healthy subjects and 20 post-AMI patients. Thereafter, a case-control approach was used to compare the prognostic significance of edited and nonedited fractal scaling exponents in a random elderly population and in a post-AMI population. Detrended fluctuation analysis (DFA) was used to measure the short-term (alpha1) and long-term (alpha2) fractal scaling exponents. RESULTS Artificial PBs caused a more pronounced reduction of alpha1 value among the post-AMI patients than the healthy subjects, for example, if > 0.25% of the beats were premature a > 25% decrease in the alpha1 was observed in post-AMI patients, but 4% of the premature beats were needed to cause a 25% reduction in alpha1 in healthy subjects. Both edited (1.01 +/- 0.31 vs 1.19 +/- 0.27, P < 0.01) and unedited alpha1 (0.71 +/- 0.33 vs 0.89 +/- 0.36, P < 0.05) differed between the patients who died (n = 42) and those who survived (n = 42) after an AMI. In the general population, only unedited alpha1 differed significantly between survivors and those who died (0.96 +/- 0.19 vs 0.83 +/- 0.27, P < 0.05). CONCLUSIONS Unedited premature beats result in an increase in the randomness of short-term R-R interval dynamics, particularly in post-AMI patients. Premature beats must not necessarily be edited when fractal analysis is used for risk stratification.

/pdf/effects-and-significance-of-premature-beats-on-fractal-2b7yuxi13t.pdf

Effects and significance of premature beats on fractal correlation properties of R-R interval dynamics.

Determination of whether results of cardiovascular reflex tests and spectral analysis of heart rate variability are age dependent and whether there is correlation between results of both, cardiovascular reflex tests (the Valsalva manoeuvre, deep breathing test, handgrip test, cold face stimulus test, orthostatic test) and spectral analysis of heart rate variability were performed on 83 healthy volunteers of both genders, aged 21 to 70 years. We found that results of all heart rate based tests and results of spectral analysis decreased with aging, while results of blood pressure based tests did not. Parasympathetic activity predominated in younger subjects, while in older subjects sympathetic activity was dominant. Valsalva, deep breathing, and orthostatic ratios correlated with integrals of amplitude spectra in the standing posture and deep breathing and cold face stimulus ratios with integrals of amplitude spectra in the supine posture, whereas blood pressure changes during handgrip and orthostatic test did not correlate with integrals of the amplitude spectra. These findings suggest that tests based on heart rate may be more sensitive than tests based on blood pressure changes. This study supports the use of spectral analysis as an additional clinical test of autonomic nervous system function and stresses the importance of age in the evaluation of the results of autonomic nervous system function testing.

A comparison of cardiovascular reflex tests and spectral analysis of heart rate variability in healthy subjects.

Heart rate variability (HRV) is used in the assessment of cardiovascular health. However, often contradictory results have impeded the efficient use of HRV in clinical practice. HRV signals can contain artifacts leading to errors in the interpretation of HRV results. Various methods have been used for artifact editing, but there is relatively little information on how the actual editing can influence the HRV measures. The main aim of this thesis was to improve the reliability of HRV analysis by concentrating on the HRV signal preprocessing methods. The effects of three editing methods on the HRV of short (512 R-R) and long-term (24-hour) R-R interval data were studied with non-edited and edited data from healthy subjects (n=10) and patients with acute myocardial infarction (AMI) (n=10). The effects of ectopic beats on short (α1) and long-term (α2) fractal scaling exponents were studied by inserting artificial ectopic beats into the HRV signals of 20 healthy subjects and 20 AMI patients. The prognostic significance of edited and non-edited α1 and α2 was studied in random elderly (n=84) and post-AMI (n=84) populations. A new method to quantify respiratory sinus arrhythmia (RSA) was developed based on the HRV signals of 13 healthy subjects. A new measure, the RSA index, was defined to evaluate the risk to sudden cardiac death (SCD) in 1631 AMI patients. Lastly, a new algorithm was developed in order to edit heart rate (HR) turbulence occurring immediately after a ventricular premature beat (VPB). The effects of HR turbulence editing on the HRV analysis were studied in 267 AMI patients. Editing had distinct effects on the HRV analysis depending on the editing method and data type. Deletion editing was found to be unsuitable for the HRV spectrum analysis. There was no universal editing method for the time and frequency domain HRV analyses. Unedited ectopic beats increased the randomness of short-term R-R interval dynamics, especially in AMI patients. However, unedited α1 differed significantly between survivors and those who died during the follow-up. Ectopic beats do not necessarily need to be edited if fractal analysis is used in the risk evaluation. A depressed RSA index was found to be a strong predictor of SCD but a weak predictor of non-SCD in AMI patients. Editing of HR turbulence affected differently the various HRV measures. ULF and VLF components were most clearly influenced by HR turbulence removal. The amount of VBPs had an important impact on the results. When the VBPs/hour were >50, ULF and VLF were >30% lower after turbulence removal. The results of this thesis highlight the importance of editing the erroneous or irrelevant R-R interval oscillation in an HRV analysis. The careful choice of preprocessing method is essential if one wishes to obtain reliable HRV analyses for clinical purposes.

/pdf/analysis-of-heart-rate-variability-from-24-hour-ambulatory-z6clki350v.pdf

Analysis of heart rate variability from 24-hour ambulatory electrocardiographic recordings : significance of preprocessing of R-R interval time series

The purpose of this study was to examine the effect of missing QRS waves upon the RRV power spectra and to show how the RRV spectra can be restored with a simple method for eliminating gaps in the data. >

http://ieeexplore.ieee.org/iel2/3299/9921/00470131.pdf

Techniques for enhancement of RR interval variability power spectra in short epochs of simian monitor ECGs

Short-term heart rate dynamics : methodology and novel applications

The authors investigate the effect of an artifact introduced by automated QRS complex recognition on two methods providing spectral heart rate variability (HRV) indices derived from 24-h electrocardiogram (ECG) recordings: (a) fast Fourier transformation and (b) peak-to-trough analysis. Both methods were used to measure the HRV spectra of 557 postmyocardial infarction patients before and after manual editing of the automated analysis. The low-, medium-, and high-frequency HRV components were calculated. For each method and each spectral component the results obtained from unedited and edited ECG analysis were compared. Both methods were substantially affected by the artifact. Spectral HRV analysis should be performed only on carefully verified 24-h ECG recognition. >

J. Gill

Papers

Influence of the artefact of computerised analysis of 24-hour electrocardiograms on the measurement of spectral components of heart rate variability