See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/6687018
`
`Heart rate variability: A review
`
`Article  in  Medical & Biological Engineering & Computing · January 2007
`
`DOI: 10.1007/s11517-006-0119-0 · Source: PubMed
`
`CITATIONS
`1,721
`
`5 authors, including:
`
`READS
`14,208
`
`U Rajendra Acharya
`Ngee Ann, Singapore University of Social Science, Singapore; University of Malaya…
`
`Paul Joseph
`National Institute of Technology Calicut
`
`726 PUBLICATIONS   31,467 CITATIONS   
`
`53 PUBLICATIONS   3,234 CITATIONS   
`
`SEE PROFILE
`
`Choo Min Lim
`Ngee Ann Polytechnic
`
`164 PUBLICATIONS   8,775 CITATIONS   
`
`SEE PROFILE
`
`SEE PROFILE
`
`Jasjit Suri
`
`505 PUBLICATIONS   11,024 CITATIONS   
`
`SEE PROFILE
`
`Some of the authors of this publication are also working on these related projects:
`
`Ethics of Biomaterials for Implants View project
`
`CT evaluation of intraplaque hemorrhage View project
`
`All content following this page was uploaded by U Rajendra Acharya on 26 May 2014.
`
`The user has requested enhancement of the downloaded file.
`
`APPLE 1024
`
`1
`
`

`

`Med Bio Eng Comput (2006) 44:1031–1051
`DOI 10.1007/s11517-006-0119-0
`
`R E V I E W A R T I C L E
`
`Heart rate variability: a review
`
`U. Rajendra Acharya Æ K. Paul Joseph Æ
`N. Kannathal Æ Choo Min Lim Æ Jasjit S. Suri
`
`Received: 19 October 2005 / Accepted: 10 October 2006 / Published online: 17 November 2006
`Ó International Federation for Medical and Biological Engineering 2006
`
`Abstract Heart rate variability (HRV) is a reliable
`reflection of the many physiological factors modulating
`the normal rhythm of the heart. In fact, they provide a
`powerful means of observing the interplay between the
`sympathetic and parasympathetic nervous systems. It
`shows that the structure generating the signal is not
`only simply linear, but also involves nonlinear contri-
`butions. Heart rate (HR) is a nonstationary signal; its
`variation may contain indicators of current disease, or
`warnings about impending cardiac diseases. The indi-
`cators may be present at all times or may occur at
`random—during certain intervals of the day. It is
`strenuous and time consuming to study and pinpoint
`abnormalities in voluminous data collected over sev-
`eral hours. Hence, HR variation analysis (instanta-
`neous HR against time axis) has become a popular
`noninvasive tool for assessing the activities of the
`autonomic nervous system. Computer based analytical
`tools for in-depth study of data over daylong intervals
`can be very useful in diagnostics. Therefore, the HRV
`signal parameters, extracted and analyzed using com-
`puters, are highly useful in diagnostics. In this paper,
`
`U. Rajendra Acharya (&) N. Kannathal C. M. Lim
`Department of ECE, Ngee Ann Polytechnic,
`535 Clementi Road, Singapore, Singapore 599 489
`e-mail: aru@np.edu.sg
`
`K. Paul Joseph
`Electrical Engineering, National Institute of Technology
`Calicut, Calilcut 673601 Kerala, India
`
`J. S. Suri
`Idaho’s Biomedical Research Institute, ID, USA
`
`J. S. Suri
`Biomedical Technologies Inc., Westminster, CO, USA
`
`we have discussed the various applications of HRV and
`different linear, frequency domain, wavelet domain,
`nonlinear techniques used for the analysis of the HRV.
`Keywords Heart rate variability Autonomic nervous
`system Poincare plot Surrogate data ANOVA test
`Phase space plot Correlation dimension Lyapunov
`exponent Approximate entropy Sample entropy
`Hurst exponent Wavelet transform Recurrent plot
`
`1 Introduction
`
`Heart rate variability (HRV), the variation over time
`of the period between consecutive heartbeats, is pre-
`dominantly dependent on the extrinsic regulation of
`the heart rate (HR). HRV is thought to reflect the
`heart’s ability to adapt to changing circumstances by
`detecting and quickly responding to unpredictable
`stimuli. HRV analysis is the ability to assess overall
`cardiac health and the state of the autonomic nervous
`system (ANS)
`responsible for
`regulating cardiac
`activity.
`HRV is a useful signal for understanding the status
`of the ANS. HRV refers to the variations in the beat
`intervals or correspondingly in the instantaneous HR.
`The normal variability in HR is due to autonomic
`neural regulation of the heart and the circulatory sys-
`tem [111]. The balancing action of the sympathetic
`nervous system (SNS) and parasympathetic nervous
`system (PNS) branches of the ANS controls the HR.
`Increased SNS or diminished PNS activity results in
`cardio-acceleration. Conversely, a low SNS activity or
`a high PNS activity causes cardio-deceleration. The
`degree of variability in the HR provides information
`
`123
`
`2
`
`

`

`1032
`
`Med Bio Eng Comput (2006) 44:1031–1051
`
`about the functioning of the nervous control on the HR
`and the heart’s ability to respond.
`Past 20 years have witnessed the recognition of the
`significant relationship between ANS and cardiovas-
`cular mortality including sudden death due to cardiac
`arrest [69, 115, 120]. Numerous numbers of papers
`appeared in connection with HRV related cardiologi-
`cal issues [8, 13, 57, 58, 64, 87, 97, 112] reiterates the
`significance of HRV in assessing the cardiac heath. The
`interest in the analysis of HRV (i.e., the fluctuations of
`the heart beating in time,) is not new. Furthermore,
`much progress was achieved in this field with the ad-
`vent of low cost computers with massive computational
`power, which fueled many recent advances.
`Tulen and Man in t’veld [125] have found that, HR,
`diastolic blood pressure (BP), mid-frequency band
`power of HR and systolic BP, and plasma adrenaline
`concentrations showed significant increase when changed
`from supine to sitting to standing posture. Viktor et al.
`[132] have studied the variation of HR spectrogram and
`breathing rates in lateral and supine body positions.
`Recently, new dynamic methods of HRV quantification
`have been used to uncover nonlinear fluctuations in
`HR, that are not otherwise apparent. Several methods
`have been proposed: Lyapunov exponents [105], 1/f
`slope [64], approximate entropy (ApEn) [93] and
`detrended fluctuation analysis (DFA) [91].
`Heart rate variability, i.e., the amount of HR fluctu-
`ations around the mean HR, can be used as a mirror of
`the cardiorespiratory control system. It is a valuable tool
`to investigate the sympathetic and parasympathetic
`function of the ANS. The most important application of
`HRV analysis is the surveillance of postinfarction and
`diabetic patients. HRV gives information about the
`sympathetic-parasympathetic autonomic balance and
`thus about the risk for sudden cardiac death (SCD) in
`these patients. HRV measurements are easy to perform,
`noninvasive, and have good reproducibility, if used
`under standardized conditions [46, 63].
`Kovatchev et al. [65] have introduced the sample
`asymmetry analysis (SAA) and illustrate its utility for
`assessment of HR characteristics occurring, early in the
`course of neonatal sepsis and systemic inflammatory
`response syndrome (SIRS). Compared with healthy
`infants,
`infants who experienced sepsis had similar
`sample asymmetry in health, and elevated values be-
`fore sepsis and SIRS (p = 0.002). Cysarz et al. [29]
`have demonstrated that the binary symbolization of
`RR interval dynamics, which at first glance seems to be
`an enormous waste of information, gives an important
`key to a better understanding of normal heart period
`regularity. Furthermore, differential binary symbolization
`
`123
`
`still enables the identification of nonlinear dynamical
`properties.
`Recently, Verlinde et al. [129] have compared the
`HRV of aerobic athletes with the controls and showed
`that the aerobic athletes have an increased power in all
`frequency bands. These results are in accordance with
`values obtained by spectral analysis using the Fourier
`transform, suggesting that wavelet analysis could be an
`appropriate tool to evaluate oscillating components in
`HRV. But, in addition to classic methods, it also gives a
`time resolution. Time-dependent spectral analysis of
`HRV using the wavelet transform was found to be
`valuable for explaining the patterns of cardiac rate
`control during reperfusion. In addition, examination of
`the entire record revealed epochs of markedly dimin-
`ished HRV in two patients, which attribute to vagal
`saturation [123]. A method for analyzing HRV signals
`using the wavelet transform was applied to obtain a
`time-scale representation for very low-frequency
`(VLF), low-frequency (LF) and high-frequency (HF)
`bands using the orthogonal multiresolution pyramidal
`algorithm [44]. Results suggest that wavelet analysis
`provides useful information for the assessment of dy-
`namic changes and patterns of HRV during myocardial
`ischaemia. Time-frequency parameters
`calculated
`using wavelet transform and extracted from the noc-
`turnal heart period analysis appeared as powerful tools
`for obstructive sleep apnoea syndrome diagnosis [104].
`Time-frequency domain analysis of the nocturnal HRV
`using wavelet decomposition could represent an effi-
`cient marker of obstructive sleep apnoea syndrome
`[104]. Its added ease of use and interpretation is of
`interest in considering the high prevalence of sleep-
`related breathing disorders in a general middle-aged,
`at-risk population. Recently, Schumacher et al. [114]
`have explained the use of linear and nonlinear analysis
`in the analysis of the HR signals. The affect of ANS,
`BP, myocardial infarction (MI), nervous system, age,
`gender, drugs, diabetes, renal failure, smoking, alcohol,
`sleep on the HRV are discussed in detail.
`Power spectral analysis of beat-to-beat HRV has
`provided a useful means of understanding the interplay
`between autonomic and cardiovascular functionality.
`Mager et al. [73] have developed an algorithm that
`utilizes continuous wavelet transform (CWT) para-
`meters as inputs to Kohonen’s self-organizing map
`(SOM), for providing a method of clustering subjects
`with similar wavelet transform signatures . Bracic et al.
`[18] have analyzed human blood flow in the time-fre-
`quency domain, and used the wavelet
`transform
`(Morlet) which gives good time resolution for high
`frequency components and good frequency resolution
`
`3
`
`

`

`Med Bio Eng Comput (2006) 44:1031–1051
`
`1033
`
`for LF components . Recently, HR (using Morlet
`wavelet as mother wavelet) for different cardiac ar-
`rhythmias was proposed (Oliver et al. 2004). Shimoj-
`ima et al. [118] have used Morlet mother wavelet to
`evaluate the performance of frequency power spec-
`trum during QRS in intraventricular conduction
`abnormalities (IVCA). They have observed that there
`is reduction of the low frequency power in IVCA and
`the increased power and number of peaks in high fre-
`quency range in IVCA with MI.
`
`1.1 The autonomic nervous system
`
`The ANS have sympathetic and parasympathetic
`components. Sympathetic stimulation, occurring in re-
`sponse to stress, exercise and heart disease, causes an
`increase in HR by increasing the firing rate of pace-
`maker cells in the heart’s sino-atrial node. Parasym-
`pathetic activity, primarily resulting from the function
`of internal organs, trauma, allergic reactions and the
`inhalation of irritants, decreases the firing rate of
`pacemaker cells and the HR, providing a regulatory
`balance in physiological autonomic function. The sep-
`arate rhythmic contributions from sympathetic and
`parasympathetic autonomic activity modulate the
`heart rate (RR) intervals of the QRS complex in the
`electrocardiogram (ECG), at distinct
`frequencies.
`Sympathetic activity is associated with the low fre-
`quency range (0.04–0.15 Hz) while parasympathetic
`activity is associated with the higher frequency range
`(0.15–0.4 Hz) of modulation frequencies of the HR.
`This difference in frequency ranges allows HRV
`analysis to separate sympathetic and parasympathetic
`contributions evident. This should enable preventive
`intervention at an early stage when it is most beneficial.
`
`A method to describe relationships between short-
`term BP fluctuations and heart-rate variability in rest-
`ing subjects was analyzed in the frequency domain [9,
`14–16]. Relationships between pressure and interval
`variability indicate that
`the 10-s variability, which
`indicate in systolic pressure, leads the interval variation
`by two to three beats and manifest in cross-spectra.
`However no such lag is found between the respiration-
`linked variations in systolic pressure. And later they
`[14–16] have proposed a simple model to interpret the
`results of spectral analysis of BP and HR data. The
`baroreflex equation of the model describe the data
`only in the region of respiratory frequencies. The shape
`of the phase spectrum of systolic pressures against
`intervals was modeled by difference equations. But no
`physiological interpretation of these equations was gi-
`ven. They have proved that, the spectral properties of
`the input signal can not be recovered fully from the
`interval spectrum, nor from the spectrum of counts, the
`more so as physiological series of events were not be
`generated by an ideal
`integrated pulse frequency
`modulation (IPFM) model
`[14–16]. Recently,
`the
`European Society of Hypertension working group on
`baroreflex and cardiovascular variability, in which 11
`centres participated, has produced a comprehensive
`database which is available for testing and comparison
`of methods [66]. Recently, Westerhof et al. [137] have
`proposed a cross-correlation baro-flex sensitivity
`(xBRS) technique for the computation of time-domain
`baroreflex sensitivity on spontaneous BP and HRV
`using EUROBAVAR data set. They proved that, the
`xBRS method may be considered for experimental and
`clinical use, because the values yielded were correlated
`strongly with and was close to the EUROBAVAR
`averages.
`
`1.2 HRV and blood pressure
`
`1.3 HRV and myocardial infarction
`
`Several structural and functional alterations of the
`cardiovascular system that are frequently found in
`hypertensive individuals may increase their cardiovas-
`cular risk beyond that induced by the BP elevation
`alone. Electrocardiographic evidence of left ventricular
`hypertrophy (LVH) and strain are associated with in-
`creased morbidity and mortality. HRV is significantly
`reduced in patients with LVH secondary to hyperten-
`sion or aortic valve disease. Cardiac vagal nerve
`activity is influenced by the arterial baroreflex. The
`amplitude of respiratory sinus arrhythmia (HRV) has
`been found to correlate with baroreflex sensitivity
`which is reduced in hypertension and diabetes. This
`reduction in baroreflex sensitivity is correlated with
`cardiac LVH.
`
`A predominance of sympathetic activity and reduction
`in parasympathetic cardiac control has been found in
`patients with acute MI [108]. Sympathetic activity de-
`creases the fibrillation threshold and predisposes to
`ventricular fibrillation (VF). Vagal activity increases
`the threshold and appears to protect against malignant
`ventricular tachyarrhythmias [117, 138]. The degree of
`respiratory sinus arrhythmia shows a linear relation
`with parasympathetic cardiac control [61, 75] and thus
`can be used as a prognostic tool in patients, who have
`had a MI. It was shown that, the HRV decreases with
`the recent MI [21, 22]. Despite the beneficial effects on
`clinical variables, exercise training did not markedly
`alter HRV indexes in subjects after MI [32]. A signif-
`icant decrease in SDRR and HF power in the control
`
`123
`
`4
`
`

`

`1034
`
`Med Bio Eng Comput (2006) 44:1031–1051
`
`group suggested an ongoing process of sympathovagal
`imbalance in favor of sympathetic dominance in un-
`trained patients after MI with new-onset left ventric-
`ular dysfunction.
`
`1.4 HRV and nervous system
`
`Disorders of the central and peripheral nervous system
`have effects on HRV. The vagally and sympathetically
`mediated fluctuations in HR may be independently
`affected by some disorders. All normal cyclic changes in
`HR are reduced in the presence of severe brain damage
`[71] and depression [21, 22]. HRV was less accurate
`than the Glasgow Coma Scale in predicting outcome.
`But it was easily accessible and may provide informa-
`tion about the patient’s neurologic status [67]. In serial
`determinations, the rate of return of normal HRV may
`reflect the subsequent state of neuronal function.
`The significance of HRV analysis in psychiatric
`disorders arises from the fact that one can easily detect
`a sympatho-vagal imbalance (relative cholinergic and
`adrenergic modulation of HRV), if it exists in such
`pathologies. There is conflicting reports about the
`HRV and the major depression. It is proved that, in
`physically healthy depressed adults the HRV does not
`vary from healthy subjects [113].
`
`1.5 HRV and cardiac arrhythmia
`
`A complex system like cardiovascular system cannot
`be linear in nature and by considering it as a nonlinear
`system, can lead to better understanding of the system
`dynamics. Recent
`studies have also stressed the
`importance of nonlinear techniques to study HRV in
`issues related to both health and disease. The progress
`made in the field using measures of chaos has attracted
`the scientific community to apply these tools in study-
`ing physiological systems, and HRV is no exception.
`There have been several methods of estimating invar-
`iants from nonlinear dynamical systems being reported
`in the literature. Recently, Fell et al. [39] and Rad-
`hakrishna et al. [99] have tried the nonlinear analysis
`of ECG and HRV signals, respectively. Also, Paul
`et al. (2002) showed that coordinated mechanical
`activity in the heart during VF may be made visible in
`the surface ECG using wavelet transform. Owis et al.
`[86] have used nonlinear dynamical modeling in ECG
`arrhythmia detection and classification. Acharya et al.
`[1–4] have classified the HRV signals using nonlinear
`techniques, and artificial
`intelligence into different
`groups. Dingfie et al.
`[46] have classified cardiac
`arrhythmia into six classes using autoregressive (AR)
`modeling.
`
`123
`
`1.6 HRV in diabetes
`
`Diabetes can cause severe autonomic dysfunction and
`can be responsible for several disabling symptoms,
`including SCD. Although traditional measures of
`autonomic function are able to document the presence
`of neuropathy, in general they are only abnormal when
`there is severe symptomatology. Thus by the time
`changes in function were evident, the natural course of
`autonomic neuropathy was well established. HRV and
`SCD Ventricular tachyarrhythmias represent a leading
`cause of SCD in the community.
`The pathophysiology of SCD was probably an inter-
`action between an abnormal anatomical substrate such
`as coronary artery disease with associated myocardial
`scarring, LVH or cardiomyopathy, and transient func-
`tional disturbances which trigger the terminal dys-
`rhythmia. This may include factors such as ischaemia,
`premature beats, electrolyte disturbance and fluctua-
`tions in autonomic balance.
`The decreased beat-to-beat variability during deep
`breathing in diabetic neuropathy was first reported by
`Wheeler and Watkins [139] and confirmed by many
`others [92]. In studies comparing cardiac autonomic
`function tests and HRV indices (based on both short
`(5-min) and 24-h ECG recordings), show that, in dia-
`betic patients without abnormal function tests, HRV
`was lowered [92]. It was concluded that cardiac
`(parasympathetic) autonomic activity was diminished
`in diabetic patients before clinical symptoms of neu-
`ropathy become evident [92, 119, 134].
`
`1.7 HRV and renal failure
`
`In patients with renal failure, autonomic function tests
`have been done [37, 146], followed by HRV indices [41]
`and spectral analysis of HR [10]. Although autonomic
`function tests revealed predominant impairment of the
`PNS [146], spectral analysis exhibited a strong reduction
`in the HR power spectrum at all frequency ranges, both
`sympathetically and parasympathetically [10]. The
`relationship between HRV parameters and electrolyte
`ion concentrations in both pre- and post-dialysis [124].
`The 5-min HRV of 20 chronic renal failure (CRF)
`patients were analyzed. Results revealed that calcium is
`negatively correlated to the mean of RR intervals and
`normalized HF power after hemodialysis. A model of
`baroreflex control of BP was proposed in terms of a
`delay differential equation and was used to predict the
`adaptation of short-term cardiovascular control in CRF
`patients [68]. They showed that in CRF patients, the
`mean power in the LF band was higher and lower in the
`
`5
`
`

`

`Med Bio Eng Comput (2006) 44:1031–1051
`
`1035
`
`HF bands than the corresponding values in the healthy
`subjects.
`
`1.8 HRV and gender, age
`
`It is proved that, the HRV depends on the age and sex
`also. The HRV was more in the physically active young
`and old women [30, 110]. It was proved by Emese et al.
`[80] that the alert new borns have lower HR variation
`in the boys than in the case of girls. The HR variation
`for healthy subjects from 20 to 70 years was studied by
`Bonnemeir et al. 17 and found that the HRV decreases
`with age and variation is more in the case of female
`than men.
`Previous studies have assessed gender and age-re-
`lated differences in time and frequency domain indices
`[100] and some nonlinear component of HRV. There
`also seemed to be a significant difference between day
`and night hours when studying HRV indices using
`spectral and time domain methods [100, 144].
`The amount of HRV is influenced by physiologic
`and maturational factors. Maturation of the sympa-
`thetic and vagal divisions of the ANS results in an in-
`crease in HRV with gestational age [128] and during
`early postnatal life [128]. HRV decreases with age [3,
`4]. This decline starts in childhood [116]. Infants have
`a high sympathetic activity that decreases quickly
`between ages 5 and 10 years [40]. The influence of
`provocation on HRV (i.e., standing and fixed breath-
`ing) is more pronounced at younger ages [116]. In
`adults, an attenuation of respiratory sinus arrhythmia
`with advancing age usually predominates [70, 135]. It
`was shown that compared to men, women are at lower
`risk of coronary heart disease [140].
`
`1.9 HRV and drugs
`
`Heart rate variability can be significantly influenced by
`various groups of drugs. The influence of medication
`should be considered, while interpreting HRV. On the
`other hand, HRV can be used to quantify the effects of
`certain drugs on the ANS.
`The effects of beta-blockers and calcium channel
`blockers on the heart rate variability have been studied
`in postinfarction and hypertensive patients [12, 28, 51].
`With spectral analysis, it is possible to unravel the
`sympathetic and parasympathetic activities of these
`drugs and thus explain their protective effects in car-
`diac diseases. In normotensive adults, the beta-adren-
`ergic blocker atenolol appears to augment vagally
`mediated fast fluctuations in HR [79]. Guzzetti et al.
`[51] studied the effect of atenolol
`in patients with
`essential hypertension. They found not only an increase
`
`in HF fluctuations, but also a decrease in the sympa-
`thetically mediated LF oscillations. This decrease in
`sympathetic activity was also noticed in postinfarction
`patients using metoprolol [12] and in patients with heart
`failure using acebutolol
`[28]. Thus, beta-blockers
`are able to restore the sympathetic–parasympathetic
`balance in cardiovascular disease. Effect of Omacor on
`HRV parameters in patients with recent uncomplicated
`MI was studied [88]. And the study, quantify improve-
`ment in time domain HRV indices and can assess the
`safety of administering Omacor to optimally treated
`post-infarction patients. Eryonucu et al.
`[36] have
`investigated the effects of ß2 -adrenergic agonist therapy
`on HRV in adult asthmatic patients by using frequency
`domain measures of HRV. The LF and LF/HF ratio
`increased and TP decreased at 5, 10, 15 and 20 min
`after the salbutamol and the terbutaline inhalation, HF
`will not change significantly after the salbutamol and
`terbutaline inhalation.
`
`1.10 HRV and smoking
`
`Studies have shown that smokers have increased sym-
`pathetic and reduced vagal activity as measured by
`HRV analysis. Smoking reduces the HRV. One of the
`mechanisms by which smoking impairs the cardiovas-
`cular function is its effect on ANS control [52, 72, 83].
`Altered cardiac autonomic function, assessed by
`decrements in HRV, is associated with acute exposure
`to environmental tobacco smoke (ETS) and may be
`part of the pathophysiologic mechanisms linking ETS
`exposure and increased cardiac vulnerability [98]. Re-
`cently Zeskind and Gingras [145] have shown that
`cigarette exposed fetuses have lower HRV and dis-
`rupted temporal organization of autonomic regulation
`before effects of parturition, postnatal adaptation, and
`possible nicotine withdrawal contributed to differences
`in infant neurobehavioral function. Also, it was proved
`that, the vagal modulation of the heart had blunted in
`heavy smokers, particularly during a parasympathetic
`maneuver. Blunted autonomic control of the heart may
`partly be associated with adverse event attributed to
`cigarette smoking [11].
`
`1.11 HRV and alcohol
`
`HRV reduces with the acute ingestion of alcohol,
`suggesting sympathetic activation and/or parasympa-
`thetic withdrawal. Malpas et al. [76] have demon-
`strated vagal neuropathy in men with chronic alcohol
`dependence using 24 h HRV analysis.
`Ryan et al. [109] have previously reported a strong
`positive association between average day time and
`
`123
`
`6
`
`

`

`1036
`
`Med Bio Eng Comput (2006) 44:1031–1051
`
`nighttime HR measured during 24 h ambulatory BP
`monitoring and usual alcohol intake. ECG indices of
`vagal activity have been reported to have significantly
`lower indices of cardiac vagal nerve activity than nor-
`mal volunteers, in acute alcoholic subjects [76, 90, 106].
`
`1.12 HRV and sleep
`
`The results from Togo and Yamamoto [122] suggest
`that mechanisms involving electroencephalographic
`desynchronization and/or conscious
`states of
`the
`brain are reflected in the fractal component of HRV.
`Compared to stages 2 and 4 non-REM sleep, the total
`spectrum power was significantly higher in REM sleep
`and its value gradually increased in the course of each
`REM cycle [20]. The value of the VLF component
`(reflects slow regulatory mechanisms, e.g., the renin-
`angiotensin system,
`thermoregulation) was
`signifi-
`cantly higher in REM sleep than in stages 2 and 4 of
`non-REM sleep. The LF spectral component (linked to
`the sympathetic modulation) was significantly higher in
`REM sleep than in stages 2 and 4 non-REM sleep.
`Patients with sleep apnoea tend to have a spectral peak
`lying between 0.01 and 0.05cycles/beat, with the width
`of the peak indicating variability in the recurrence rate
`of the apnoea. In most of the subjects, the frequency
`spectrum immediately below the apnoea peak was
`relatively flat. The first visual analysis of the single
`computed spectrum from each subject led to a correct
`classification score of 28/30 (93%) [31]. Gates et al.
`[45] suggested that, a long-lasting alterations existed in
`autonomic function in snoring subjects.
`
`1.13 In infants
`
`Investigations in the fetus and newborn have revealed
`that, during rapid eye movement (REM) sleep long-
`term variability (LTV) was increased and short-term
`variability (STV) decreased compared to during non-
`REM sleep [107, 128]. These differences between
`REM and non-REM sleep, were due to a shift in the
`vagal-sympathetic balance from a higher sympathetic
`tone during REM sleep to a higher vagal tone during
`non-REM sleep [127]. In addition, the slower and more
`regular breathing in non-REM sleep (more respiratory
`sinus arrhythmia, thus more STV) contributes to the
`
`differences found. Higher HR and lower total power
`lower was found in infants than in children [133]. It was
`proved that, during the deep sleep, the HRV decreases
`in healthy subjects [34]. HF power was higher in chil-
`dren than in infants. In infants and children, the ratio
`between LF and HF powers changed with the various
`sleep stages (p < 0.02 in infants; p < 0.01 in children):
`it decreased during deep sleep and increased during
`REM sleep.
`Heart rate variability is a measure of variations in
`the HR. Figure 1 shows the variation of the HR of a
`normal subject. It is usually calculated by analyzing the
`time series of beat-to-beat intervals from ECG or
`arterial pressure tracings. Various measures of HRV
`have been proposed, which can roughly be subdivided
`into time domain, frequency domain and nonlinear
`domain measures.
`
`2 Methods
`
`2.1 Time domain analysis
`
`Two types of HRV indices are distinguished in time
`domain analysis. Beat-to-beat or STV indices represent
`fast changes in HR. LTV indices are slower fluctua-
`tions (fewer than 6 min–1). Both types of indices are
`calculated from the RR intervals occurring in a chosen
`time window (usually between 0.5 and 5 min). From
`the original RR intervals, a number of parameters are
`calculated: SDNN, the standard deviation of the NN
`intervals, SENN is the standard error, or standard error
`of the mean, is an estimate of the standard deviation of
`the sampling distribution of means, based on the data,
`SDSD is the standard deviation of differences between
`adjacent NN intervals. RMSSD, the root mean square
`successive difference of intervals, pNN50%, the num-
`ber of successive difference of intervals which differ by
`more than 50 ms expressed as a percentage of the total
`number of ECG cycles analyzed. The statistical
`parameters SDNN, SENN, SDSD, RMSSD, NN50
`(%), and pNN50% [120] can be used as time domain
`parameters.
`The statistical parameters SDNN, SENN, SDSD,
`RMSSD, pNN50% and TINN are found to have bigger
`value for the classes like preventricular contraction
`
`Fig. 1 Heart rate variation of
`a normal subject
`
`123
`
`7
`
`

`

`Med Bio Eng Comput (2006) 44:1031–1051
`
`1037
`
`(PVC), sick sinus syndrome (SSS), and atrial fibrilla-
`tion (AF) due to higher RR variation. And for the
`slowly varying signal like complete heart block (CHB),
`left bundle branch block (LBBB) and Ischemic/dilated
`cardiomyopathy, these parameters will be lesser be-
`cause of the smaller RR variation [1, 3].
`
`2.1.1 Analysis by geometrical method
`
`Geometrical methods present RR intervals in geo-
`metric patterns and various approaches have been
`used to derive measures of HRV from them. The
`triangular index is a measure, where the length of RR
`intervals serves as the x-axis of the plot and the
`number of each RR interval length serves as the y-
`axis. The length of the base of the triangle is used and
`approximated by the main peak of the RR interval
`frequency distribution diagram. The triangular inter-
`polation of NN interval histogram (TINN) is the
`baseline width of the distribution measured as a base
`of a triangle, approximating the NN interval distribu-
`tion (the minimum of HRV). Triangular interpolation
`approximates the RR interval distribution by a linear
`function and the baseline width of this approximation
`triangle is used as a measure of the HRV index [38,
`74]. This triangular index had a high correlation with
`the standard deviation of all RR intervals. But it is
`highly insensitive to artifacts and ectopic beats, be-
`cause they are left outside the triangle. This reduces
`the need for preprocessing of the recorded data [74].
`The major advantage of geometric methods lies in
`their relative insensitivity to the analytical quality of
`the series of NN intervals. The typical values of dif-
`ferent statistical and geometric parameters of HR
`signal (Fig. 1) is shown in Table 1.
`
`2.1.2 Poincare geometry
`
`The Poincare plot, is a technique taken from nonlinear
`dynamics and portrays the nature of RR interval fluc-
`tuations. It is a plot in which each RR interval plotted
`as a function of the previous RR interval. Poincare plot
`analysis is an emerging quantitative-visual technique,
`whereby the shape of the plot is categorized into
`functional classes, that indicate the degree of the heart
`failure in a subject [143]. The plot provides summary
`information as well as detailed beat-to-beat informa-
`tion on the behavior of the heart [59]. This plot may be
`analyzed quantitatively by calculating the standard
`deviations of the distances of the R–R(i) to the lines
`y = x and y = –x + 2R–Rm, where R–Rm is the mean of
`all R–R(i) [126]. The standard deviations are referred
`
`Table 1 Result of statistical and geometric parameters of heart
`rate (Fig. 1)
`
`Time domain statistics
`
`Variable
`
`Statistical measures
`SDNN
`SENN
`SDSD
`RMSSD
`NN50
`pNN50
`Geometric measures
`RR triangular index
`TINN
`
`Units
`
`ms
`ms
`ms
`ms
`Count
`%
`
`ms
`
`Value
`
`30.0
`4.12
`36.6
`33.3
`0
`0.0
`
`0.011
`20.0
`
`to as SD1 and SD2, respectively. SD1 related to the
`fast beat-to-beat variability in the data, while SD2
`describes the longer-term variability of R–R(i) [126].
`The ratio SD1/SD2 may also be computed to describe
`the relationship between these components. Figure 2
`shows the Poincare plot of a normal subject (shown in
`Fig. 1).
`SD1/SD2 shows the ratio of short interval variation
`to the long interval variation. This ratio is more in the
`case of PVC, AF, SSS and VF due to more variation in
`the RR interval. But, this ratio falls (below no