ECG Time Series Variability Analysis

Engineering and Medicine

Herbert F. Jelinek, David J. Cornforth, Ahsan H. Khandoker, Herbert F. Jelinek, David J. Cornforth, Ahsan H. Khandoker

• Publisher: CRC Press
• Year: 2017
• Print ISBN: 9780367870157
• eBook ISBN: 9781315355610
• Topic: Mathematics
• Subtopic: Arithmetic
• Index: Mathematics

Divided roughly into two sections, this book provides a brief history of the development of ECG along with heart rate variability (HRV) algorithms and the engineering innovations over the last decade in this area. It reviews clinical research, presents an overview of the clinical field, and the importance of heart rate variability in diagnosis. The book then discusses the use of particular ECG and HRV algorithms in the context of clinical applications.

1

Introduction to ECG Time Series Variability Analysis: A Simple Overview

Herbert F. Jelinek, David J. Cornforth, and Ahsan H. Khandoker

CONTENTS

1.1 Preliminary Considerations When Measuring HRV

1.2 HRV Methods: A Short Introduction

1.3 Time Domain Measures of HRV

1.4 Frequency Domain Measures of HRV

1.5 Nonlinear Measures of HRV

1.6 Clinical Utility of HRV

References

Physiological rhythms or oscillations are the manifestation of a complex physiological system. The clinical community has long recognized that alterations in physiological rhythms are associated with disease and therefore have clinical value. Oscillations in cardiovascular systems are reflected in electrocardiogram (ECG) time series variability. For example, beat to beat variability in heart rate or heart rate variability (HRV) analysis has experienced a tremendous increase in interest from both the engineering community and medical profession, as well as from the social science, economic, and health sectors. What follows is a brief overview of the chapters included in this book, noting that each chapter was a team effort by the various laboratories around the globe that work in this field. This book is organized to provide a historical overview of the domain by Andreas Voss in Chapter 2 and a basic overview of HRV analysis and review of the basics of biosignal processing by Dragana Bajić and her coauthors in Chapter 3. Chapter 3 is aimed at readers who are new to this field or who need an overview of the basic concepts. From these introductory chapters, the book moves on to provide some groundbreaking computational applications by Gaetano Valenza and colleagues (Chapter 4) as well as the laboratory of Alberto Porta and colleagues in Chapter 5. Danuta Makowiec and coauthors discuss how graph theory may be applied to HRV analysis in Chapter 6. Many of these applications require on-site coding and Mika Tarvainen introduces Kubios in Chapter 7, which is a shareware program available from the World Wide Web that provides the opportunity to investigate biosignals processing and obtain the fundamental time and frequency domain measures as well as some nonlinear attributes of the biosignals. This software includes preprocessing options and time and frequency domain analysis as well as nonlinear HRV analysis options, for those that require a user-friendly application for HRV analysis. The remainder of the book then concentrates on several areas of clinical applications with the aim to introduce the reader to the utility of HRV. In some cases, other biosignal variability analysis methods are discussed, such as blood pressure and electroencephalogram (EEG) analysis, which can be coupled to heart rate tachograms. An important aspect of the clinical chapters is the inclusion by the authors of explanations of why they used the algorithms and they also propose more advanced methods that address the research problem better.

Thus, in Chapter 8, David Cornforth and Herbert Jelinek ask the question of how complexity measures deepen our understanding of pathophysiological processes associated with cardiac rhythm. Chapter 9, by Tatjana Lončar-Turukalo et al., is the first chapter to address biosignal coupling between blood pressure and HRV. Chandan Karmakar and coauthors then take the reader, in Chapter 10, back to a fundamental aspect of heart rate and its variability by discussing the tone-entropy feature at multiple scales.

This book does not only address how to classify or identify cardiac rhythm pathology but also covers how HRV can be used to assess the effects of training in sport and as a means of staying healthy, which is discussed in Chapter 11 by Kuno Hottenrott and Olaf Hoos. Using HRV to assess the patient response to a virtual reality neurological rehabilitation is the subject of Chapter 12, by Herbert Jelinek et al., while HRV compared to traditional outcome measures in cardiac rehabilitation is covered by Hosen Kiat’s group in Chapter 13. In Chapter 14, Ian Baguley and Melissa Nott examine changes in autonomic nervous system function in acute brain injury. Chapters 15 by Andrew Kemp and Daniel Quintana and Chapter 16 by Karl-Jürgen Bär and Andreas Voss discuss psychiatric disorders and HRV. Ahsan Khandoker, in Chapter 17, presents the recent progress in fetal ECG and fetal HRV technique. Chapter 18, by Janice Russell and Ian Spence, reviews HRV analysis in anorexia nervosa and eating disorders in general. In Chapter 19, Juha Perkiömäki and Heikki Huikuri discuss applying HRV in clinical practice following an acute myocardial infarction. Matthias Baumert outlines HRV analysis in cardiac control during normal and hypertensive pregnancy (Chapter 20). Jaqueline Phillips and Cara Hildreth then introduce, in Chapter 21, telemetry use in animal models of kidney disease. The last chapter then reaches the cellular level and investigates beat-to-beat variability in cardiomyocytes covered by Helmut Ahammer and colleagues from Graz.

However, before any biosignal analysis takes place, a number of issues have to be considered, which are briefly outlined below.

1.1 Preliminary Considerations When Measuring HRV

Methodological considerations form the crux of any research as they are a big part of using HRV as a tool in clinical practice. The number of methods proposed over the last 50 years has risen dramatically as our understanding of the physiology and pathophysiology of cardiac rhythm has grown. Time domain, frequency domain, and nonlinear methods of HRV analysis have to be chosen carefully depending on the information about the biosignal that is required.

A standard ECG signal is shown in Figure 1.1. This type of signal has been exhaustively studied and the diagnostic value of the different features is well established. The QRS complex, with R being the peak of the wave or fiducial point, is used as a surrogate point to the p-wave peak in determining the interbeat time for HRV analysis.

For HRV analysis, several preprocessing considerations have to be met. Noises in the recording and ectopic beats have to be removed. How do we deal with removed or missing beats? Manual selection of noise and ectopics is time consuming and also less likely to lead to identical outcomes if repeated. Therefore, automated preprocessing algorithms have been proposed (Marzbanrad et al. 2013; Karlsson et al. 2012; Kim et al. 2009; Thuraisingham 2006; Wessel et al. 2000; Sapoznikov et al. 1992). RR intervals are events that are not evenly spaced and therefore for some HRV analysis, especially frequency domain analysis, resampling is required and consequently the resampling frequency becomes important (Clifford and Tarassenko 2005; Struzik and Hayano 2006; Moody 1993). Current algorithms for HRV analysis tend to be applied to tachograms with reduced sampling frequency in order to minimize the data size, increase analysis speed, or as a prerequisite for evenly distributed data, while retaining high clinical accuracy (Grant et al. 2011). Resampling frequencies that have been applied for HRV analysis vary between 1 and 10 Hz with low sampling frequencies possibly leading to a loss of information. In addition, the resampling frequency also affects the HRV results in particular frequency domain measures (Singh et al. 2004). Choosing an appropriate resampling frequency is not only a function of the Nyquist frequency of the signal of interest but also of the HRV analysis employed (Abubaker et al. 2014). Within the context of preprocessing and resampling, the length of the recording also needs to be considered. In clinical practice, 10-second, 12-lead ECGs are routinely recorded in addition to longer Holter recordings, which are usually recorded for between 24 and 72 hours. However, recording lengths of 2, 5, 10, 20, or 30 minutes as well as 2 hours are not uncommon and are often a function of the HRV method used (Smith et al. 2013; Kemp et al. 2012; Grant et al. 2011; Dekker et al. 2000; de Bruyne et al. 1999; Sinnreich et al. 1998; Saul et al. 1988). HRV algorithms such as very low frequency power (VLF) or approximate entropy (ApEn) may not be suitable for use with very short recording periods although when applying even these in clinical practice, they may be sufficiently robust to provide useful information for the clinician (Jelinek et al. 2014). Teich et al. have shown that some measures provide reliable results using recordings of only a few minutes (Teich et al. 2001). In addition, when comparing HRV results, recording lengths need to be of the same duration. Automated preprocessing to remove noise and ectopic beats, resampling, and consideration of length of recording all play an important role in obtaining meaningful results in clinical practice and research. Finally, testing for stationarity is a step often neglected. Biological signals are inherently nonstationary and measures such as the correlation dimension or power spectral analysis are strongly influenced by the nonstationarity of the signal. To address this point rather than determining the extent of nonstationarity, it is suggested that when applying the power spectral analysis using a fast Fourier transform, 5-minute segments are analyzed and averaged to avoid nonstationarity features. One reason for this is that there is currently a lack of understanding relating to what constitutes too much nonstationarity when applying HRV measures that are sensitive to this characteristic of biosignals. One solution is to divide a tachogram into segments and determine the average of the segments. A measure of nonstationarity is a large standard deviation difference between the two segments (Palazzolo et al. 1998; Gao et al. 2013; Camargo et al. 2013; Pacheco et al. 2012; Chen et al. 2002; Źebrowski et al. 1999; Lempel and Ziv 1976).

FIGURE 1.1
Normal ECG signal showing the RR interval.

1.2 HRV Methods: A Short Introduction

The interval between successive R peaks is known as the RR interval (inverse of heart rate). RR intervals are obtained from the recorded ECG and the RR variation can be subjected to further analysis through a variety of algorithms in ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Contents
• Preface
• Editors
• Contributors
• 1. Introduction to ECG Time Series Variability Analysis: A Simple Overview
• 2. Historical Development of HRV Analysis
• 3. A Descriptive Approach to Signal Processing
• 4. Linear and Nonlinear Parametric Models in Heart Rate Variability Analysis
• 5. Assessing Complexity and Causality in Heart Period Variability through a Model-Free Data-Driven Multivariate Approach
• 6. Visualization of Short-Term Heart Period Variability with Network Tools as a Method for Quantifying Autonomic Drive
• 7. Analysis and Preprocessing of HRV—Kubios HRV Software
• 8. Multiscale Complexity Measures of Heart Rate Variability—A Window on the Autonomic Nervous System Function
• 9. BP and HR Interactions: Assessment of Spontaneous Baroreceptor Reflex Sensitivity
• 10. Tone–Entropy Analysis of Heart Rate Variability in Cardiac Autonomic Neuropathy
• 11. Heart Rate Variability Analysis in Exercise Physiology
• 12. Monitoring Patients during Neurorehabilitation Following Central or Peripheral Nervous System Injury: Dynamic Difficulty Adaptation
• 13. Heart Rate Variability as a Useful Parameter in Assessment of Cardiac Rehabilitation Outcome
• 14. Acquired Brain Injury Rehabilitation: What Can HRV Tell You?
• 15. Heart Rate Variability in Psychiatric Disorders, Methodological Considerations, and Recommendations for Future Research
• 16. Cardiac Autonomic Dysfunction in Patients with Schizophrenia and Their Healthy Relatives
• 17. Fetal Heart Rate Variability
• 18. Heart Rate Variability and Eating Disorders
• 19. Applying Heart Rate Variability in Clinical Practice Following Acute Myocardial Infarction
• 20. Beat-to-Beat QT Interval Variability and Autonomic Activity
• 21. The Predictive Utility of Heart Rate Variability in Chronic Kidney Disease: A Marker of Patient Outcomes
• 22. Beat-to-Beat Variability of Cardiomyocytes
• 23. Associations between Genetic Polymorphisms and Heart Rate Variability
• Index