Parametric estimation of sample entropy in heart rate variability analysis
Introduction
Heart rate variability (HRV) analysis is an important tool for evaluating cardiac autonomic regulation [1]. About 30 years ago, Pincus [2] developed a family of statistics, called approximate entropy (ApEn), to measure series regularity. Many potential applications [3], [4], [5] of this method can be found in medical research literature, in particular for detecting and testing the regularity of HRV data. Successively, to address some manifest limitations of ApEn (Pincus himself [6] reported ApEn to be a biased statistic), Richman and Moorman [7] introduced Sample Entropy (SampEn), which was also applied successfully to a wide range of problems [8], [9], [10].
In the last 20 years [11], ApEn and SampEn had been the most commonly used methods to quantify the regularity of biological data. Both metrics estimate the differential entropy rate of the series [12], [13]. However, SampEn: (i) is less prone to practical inconsistency, as it requires less lengthy series to converge to the final value and (ii) is relatively less biased even for not-so-long series [7]. Regarding the term “inconsistency”, Pincus [12] considered the problem of assessing if a stochastic process A was more regular than process B, by means of computing ApEn. He defined “consistent” those processes for which ApEn of A was always larger (or smaller) than ApEn of B, for any value of the parameters on which the metric depends, (i.e., m and r, see Section 2.1). Here we use the term “practical consistency” to refer to the fact that ApEn (or SampEn) of series SA is larger than ApEn (or SampEn) of series SB for a broad range of the parameters values.
ApEn and SampEn are extremely sensitive to data length (N), particularly for very short data sets, i.e., N ≤ 200 [11]. Hence, their estimates may be far away from what expected using longer series. Unfortunately, short series are generally used in real applications. So, issues of convergence may appear when estimating the regularity of short data. A related problem arises in spectral analysis, where long time stationary series are required to achieve lower variance of the estimates. Hence, since the works of Bishop [14] and Kay and Marple [15], parametric spectrum analysis is commonly performed on short RR series [1] of 3–5 min, which are reasonably stationary. We therefore verify in this work, if, in analogous circumstances, a SampEn computation based on a parametric representation of the series might convey new information.
The first step of a parametric approach is to select the most appropriate family of model. For HRV signal, the most commonly used models are moving average (MA), autoregressive (AR), and autoregressive moving average (ARMA). Identification of AR models has been explored largely in the literature, it requires solving simpler equations than those required for ARMA models, and AR models are maximum-entropy models (among those sharing the same autocorrelation function). Thus, AR models are mostly used for HRV maximum-entropy spectral estimation.
In this work, we will first discuss the conditions under which a parametric estimation of SampEn (and ApEn) is possible. We will limit our attention to linear AR models. Pincus [2] and then Lake [13] already tackled the problem of deriving analytical formulas of ApEn and SampEn for an AR process. Following the suggestion in [2], in this work, we have first extended the analytical expression of ApEn. Then, we have also derived an analytical expression for SampEn and tested these predictions on simulated series and real HRV data, obtained from Holter recordings.
Section snippets
ApEn
ApEn measures the likelihood that runs of patterns that are close remain close at the next incremental comparisons. The determination of this statistic is dependent on the prior selection of two unknown parameters: the length (m) of compared runs, also called templates, constructed from the series and a filtering threshold (r), i.e., the tolerance of mismatch between the corresponding elements of the templates.
Given a time series {u[i]|1 ≤ i ≤ N} of N data points, the calculation of ApEn [2] is as
Dataset
To investigate if a parametric approach is sensible in practical terms, Physionet's normal sinus rhythm (nsrdb & nsr2db), and congestive heart failure (chf2db) databases were considered for this study. The “nsrdb” database consists of 18 (5 men, age: 26–45, and 13 women, age: 20–50) long-term ECG recordings of subjects without any significant arrhythmias. “nsr2db” includes beat annotation files for 54 Holter recordings of subjects in normal sinus rhythm (30 men, age range 28.5–76, and 24 women,
Discussion
In this study, we set up a theoretical basis for the parametric estimation of sample entropy through autoregressive models, which are popularly used for power spectrum analysis of time-series (and HRV in particular). We started from the results available and developed new theoretical ones. In particular, we proposed a way to verify if series generated from an AR model are indistinguishable from the original sequence, in term of SampEn (and ApEn). The theoretical results matched the outcomes of
Conclusions
A parametric computation of SampEn through AR model is largely possible when the series are short and free of artifacts. However, sample entropy of an AR process is directly related to its autocorrelation function. Thus when estimates of SampEn, computed numerically or parametrically do agree, SampEn is mainly influenced by linear properties of the series. On the other hand, when they do not match, SampEn is truly offering information not readily available with traditional temporal and spectral
References (16)
Heart rate variability standards of measurement, physiological interpretation, and clinical use
Circulation
(1996)Approximate entropy as a measure of system complexity
Proc. Natl. Acad. Sci. U. S. A.
(1991)- et al.
Regularity patterns in heart rate variability signal: the approximate entropy approach
- et al.
Altered complexity and correlation properties of R–R interval dynamics before the spontaneous onset of paroxysmal atrial fibrillation
J. Am. Heart Assoc.
(1999) Characterization of fetal heart rate using approximate entropy
Computers in Cardiology
(2005)- et al.
Physiological time-series analysis: what does regularity quantify
Am. J. Physiol. Heart Circ. Physiol.
(1994) - et al.
Physiological time series analysis using approximate entropy and sample entropy
Am. J. Physiol. Heart Circ. Physiol.
(2000) - et al.
Sample entropy analysis of neonatal heart rate variability
Am. J. Physiol. Regul. Integr. Comp. Physiol.
(2002)
Cited by (50)
Signal processing for cardiovascular applications in p-health
2022, Personalized Health Systems for Cardiovascular DiseaseMaximum likelihood autoregressive model parameter estimation with noise corrupted independent snapshots
2021, Signal ProcessingCitation Excerpt :Autoregressive (AR) modeling of random signals is being used in many areas associated with the statistical signal processing [1] such as radar signal processing, speech processing and biomedical signal processing [2–5].
A novelty detection scheme for rolling bearing based on multiscale fuzzy distribution entropy and hybrid kernel convex hull approximation
2020, Measurement: Journal of the International Measurement ConfederationEvaluation of the Tinetti score and fall risk assessment via accelerometry-based movement analysis
2019, Artificial Intelligence in MedicineDepression recognition according to heart rate variability using Bayesian Networks
2017, Journal of Psychiatric ResearchCitation Excerpt :SD2 measures the length of the Poincaré plot, reflecting the long-term HRV. Sample entropy is usually used to calculate HRV regularity (Aktaruzzaman and Sassi, 2014; Al-Angari and Sahakian, 2007). Surveys have explored the relationship between the HRV regularity measured by the sample entropy and the dysfunction of some cardiac ANS, indicating that cardiac autonomic impairment relates to increased HRV irregularity (Weippert et al., 2014; Zhao et al., 2015).