Ebook: Frontiers of Blood Pressure and Heart Rate Analysis
An increasing number of studies indicate that the analysis of blood pressure and heart rate variability may be a valuable tool for the investigation of the mechanisms responsible for cardiovascular regulation in physiological and pathological conditions. The reader can find in the first part of this book an updated review of the techniques currently employed for the computer analysis of these signals with a particular attention to the most innovative approaches based on the non-linear analysis (including applications of the chaos theory, fractal analysis, I/f modelling) and the time-variant estimation of BP and HR characteristics. The biological interpretation of the results obtained by these computerized procedures and the applicability of these techniques in a clinical setting are fully addressed in the second part of the book.
In recent years, computer analysis of blood pressure and heart rate variability has allowed the identification of specific patterns in the fluctuations of these signals which reflect the effects of individual mechanisms involved in cardiovascular regulation. Based on the automatic assessment of these patterns, new and sensitive tools for evaluating the features of cardiovascular control have been developed. The application of these tools has led to a deeper understanding of cardiovascular regulation in daily life and to a quantitative assessment of the alterations in cardiovascular control mechanisms which may occur with ageing and in a variety of pathologic conditions.
Moreover, available experiment data seem also to support the clinical importance of the analysis of blood pressure and heart rate variability. Evidence has been provided that blood pressure and heart rate variability may have a prognostic value in cardiovascular disease such as acute myocardial infarction, congestive heart failure, arterial hypertension, diabetes mellitus and primary autonomic dysfunction.
Some of the above mentioned topics were addressed in two previous books published by IOS Press in this series. However, ongoing progress which characterises this field necessitates continuous updates in the methodology of blood pressure and heart rate signal analysis as well as in the biological interpretation of the results.
The aim of the present book is to review recent advancements in the development of new techniques for blood pressure and heart rate signal analysis, in the understanding of the physiological mechanisms responsible for heart rate and blood pressure variability, and, last but not least, in the clinical application of these techniques.
This volume includes contributions by leading experts in this area who took part in a dedicated workshop held in Florence on May 1995. We are confident that both the investigators and the clinicians working in this challenging field will find this book to be a useful reference for their professional activity.
M. Di Rienzo
Under healthy conditions, the normal cardiac (sinus) interbeat interval fluctuates in a complex manner. Quantitative analysis using techniques adapted from statistical physics reveals the presence of long-range power-law correlations extending over thousands of heartbeats. This scale-invariant (fractal) behavior suggests that the regulatory system generating these fluctuations is operating far from equilibrium. In contrast, we find that for subjects at high risk of sudden death (e.g. congestive heart failure patients) these long-range correlations break down. Application of fractal scaling analysis and related techniques provides new approaches to assessing cardiac risk and forecasting sudden cardiac death, as well as motivating development of novel physiological models of systems that appear to be “hetero-dynamic” rather than “homeo-static.”
“Surrogate data” is the basis for a technique for testing a time series for nonlinear dynamics and process nonstationarities. The theory behind surrogate data is briefly described, along with an algorithm for generating it. Examples are given of its use in detecting nonlinearity in heart rate signals, detecting non-stationarity, and estimating the sampling distribution of complicated statistics of heart rate variability.
The observation of biological systems suggests the hypothesis that nonlinear mechanisms could be involved in the control of their functions. The analysis of cardiovascular system, starting from the measurement of its state variables, seems to confirm the nonlinear nature of the control mechanisms and the presence of fractal structures in those signals. The goal of this study is to verify if a physiological control system is able to generate complex and also chaotic dynamics when periodically forced by a sinusoidal input at different frequencies. The paper analyzes a simple physiological model which accounts for the oscillations in the arterial blood pressure signal generated by the action of the baroreceptive control. The model was proposed by Kitney in 1979 and it considers the effect of the respiration signal like an external periodically forcing term. Using this model, a variety of nonlinear behaviors like the frequency entrainment, the phase locking and the frequency shift can be reproduced in different experimental situations. A study of the dynamics of the baroreceptive model through a structural stability analysis is proposed. The bifurcation diagrams classifies the different dynamical behavior of the model for different values of respiratory frequency and gain of baroreceptive system parameters. Other model parameters are fixed at realistic values. The large number of bifurcations of different types indicate that the dynamics of the model can be very complex. In fact, for values of parameters in physiological range, multiplicity of attractors, subharmonics of various periods, period doublings, quasiperiodic solutions and strange attractors get up. Results are in agreement with the hypothesis that a nonlinear dynamic model underlines the variability control.
The first observations of a 1/f trend in the spectrum of heart rate (HR) and blood pressure (BP) signals date back to 1982 and 1990, respectively. Ever since, a number of studies have suggested that the analysis of this spectral trend may be a valuable tool for investigating the physiological mechanisms involved in cardiovascular regulation and for assessing their derangement in pathological conditions. The spectral analysis of BP and HR tracings recorded in cats before and after the surgical opening of the baroreflex loop, offered us the opportunity to address some yet unclear aspects related to the 1/f trend occurring in BP and HR spectra, namely: 1) the role played by the baroreflex in the genesis of the 1/f trend; 2) the ability of a single 1/f curve to correctly model the spectra; 3) the methodology required for the estimation of the regression line corresponding to the 1/f curve when spectra are plotted in a log-Iog scale. The results we obtained indicate that 1) baroreflex denervation disrupts the 1/f trend of BP spectra but does not modify the same trend in the HR spectra; 2) in pathological conditions a single 1/f line may be inadequate to describe the BP spectra; and 3) the slope of the regression line representing the 1/f trend strongly dependent on the frequency region over which the line is estimated, thus prompting for the establishment of standards in the modeling procedure.
Beat-by-beat variations of human heartbeat intervals (heart rate variability; HRV) contain fractal components with broad band, 1/fß spectra. In this review, we first revisited the topics “what is the fractal component?” and “why should it be extracted?” A new method called coarse graining spectral analysis (CGSA) was then introduced to extract the fractal components from HRV signal. This was followed by the discussion on the accuracy aspects of CGSA and the application to the actual HRV data.
This paper introduces time domain methods for relating fluctuations in heart rate (HR) to respiration. Three methods: (1) linear stationary, (2) linear non-stationary and (3) non-linear approaches are described. (1) Experimental and signal processing contrivances called random breathing and noise injection methods enable the stable estimation of the impulse response of the HR to the instantaneous lung volume (ILV) signals. The comparison of the HR power spectra and the inpulse response revealed that the respiratory HR fluctuations are mostly affected by the parasympathetic autononic nervous activity. (2)Adaptive signal processing techniques enable recursive estimation of the impulse response, which allows us to keep track its temporal changes. Computer simulation confirmed the feasibility of the method in a practical situation. (3)An index called the degree of non-linearity (d.n.) has been introduced to evaluate how much non-linear the system is. Computer simulation confirmed the validity of the estimation method. Real data analysis showed that 30-35 % of the HR fluctuations associated with ILV cannot be predicted by the linear combination of ILV signals.
Mutual interactions between heart rate, arterial blood pressure, and instantaneous lung volume have been previously described by parametric bivariate models, which only take into account two signals at a time. An expanded trivariate autoregressive model is introduced which is able to identify simultaneously all of the possible interactions among three signals, computing the transfer functions between each pair of signals. In addition, the model computes the autospectra and its parameters for each of the three signals. For comparison, simulations using simple sinusoidal signals with superimposed white noise and an example from a gradual tilting protocol with a random interval breathing technique were analyzed with both the bivariate and the trivariate techniques. As expected, differences were found between the bivariate and trivariate technique, for the gain values of the transfer function blocks, possibly because the bivariate technique is theoretically not able to separate the true interaction between two of the signals from the contributions of the third. Further application of the trivariate expanded technique could lead to a better understanding of specific physiological mechanisms or particular pathological and non-pathological phenomena occurring in the cardiovascular control system.
Heart rate variability is currently analysed through the power spectral density (PSD) of the time series of RR interval (tachogram), estimated by either Fourier transform based or parametric methods. The hypothesis of stationarity of the tachogram, implicit in these methods, cannot be assumed when the series are derived during autonomic tests.
The PSD estimation based on the Wigner-Ville time frequency distribution (WD) shows the best performance, in terms of time and frequency resolution, among the various techniques that have been devised for computing time dependent PSD estimates. Such a technique has been applied to the analysis of tachograms, derived from multichannel recordings in normal, diabetic and Chagasic individuals stimulated by autonomic tests. The chosen tests were able of eliciting an increase in vagal tone (Valsalva manoeuvre and phenylephrine), or a sympathetic activation and a parasympathetic withdrawal (head-up tilt). The cross terms artefact associated with the WV technique has been effectively reduced by the smoothed WV estimate, trading off artefacts with a slight decrease of time resolution. The spectra of the tachograms obtained by this procedure allowed to achieve an unique characterisation of the evolving tachogram variability, with high time and frequency resolution, in all the autonomic tests.
Fluctuations of heart rate and blood pressure reflect dynamic control over tonic neuronal activity. A number of confounding factors have to be considered before cardiovascular rhythms can be used as a reliable index of autonomic nervous system integrity. The temporal profile of nonrespiratory oscillations (0.01 to 0.15 Hz with a dominant ≈0.05 Hz rhythm) (NONRF) and oscillations at respiratory frequencies (RF) (0.01 to 0.5 Hz) were analyzed by modified Wigner distribution (WD). Both RF and NONRF were continuously present during supine rest and in the upright position in normal young and elderly subjects. However, the pattern of NONRF oscillations in BP and R-R interval (RRI) becomes unstable in orthostatically intolerant patients, e.g. with vasodepressor syncope or orthostatically intolerant patients. Typically, large transient NONRF oscillations in BP occur after tilt-up and disappear with impending syncope. In patients with severe symptoms of postural tachycardia, both NONRF and RF in RRI are reduced. Loss of rhythmicity over the whole range from 0.01 to 0.5 Hz is associated with severe central lesions (like in brain stem injury). The profile of cardiovascular rhythms provides information about the integrity of the autonomic neural pathways. It could provide insights into the timing and processing of afferent impulses in the brainstem network.
The cardiovascular reactivation to a recently described videogame task i.e. a maze test was evaluated in the frequency-domain using finger blood pressure (BP) measurement in 27 untreated subjects including 5 mild hypertensives. Age ranged from 23 to 52. BP was measured with a Finapres device. Breathing was analyzed by respiratory inductance plethysmography with a Respitrace. The test consisted of a series of video displayed mazes which had to be solved within short periods of time. The recording session was divided into resting, test and recovery periods. A detrending procedure was applied to each recording prior to the fast Fourier transform (FFT), using a moving average over 20 sec. The mid-frequency (MF, 0.1 Hz) component was increased during the test for SBP and HR as well. Since a large variability was observed between individuals, we calculated the relative changes for this MF component. Increases of 33 % and 54 % in the SBP and HR MF components were observed. These MF changes could depend on the autonomic activation occuring during the performance of the mental task. In addition, the test determined a reduction of the absolute systolic BP (SBP) and heart rate (HR) high-frequency (HF, respiratory-linked) components. However these changes were associated with a modified breathing pattern characterized by a moderate tachypnea and a reduced breathing amplitude (−14%). A statistical adjustment of HF components for breathing amplitude was used to test whether changes in these HF components were due to respiration or to specific mechanisms. Therefore, the relative changes in the SBP or HR HF components were divided by the relative changes in the respiratory amplitude. This procedure demonstrated the SBP or HR HF changes during stress were entirely due to respiration since no significant changes in the normalized indices were detected during the test. This study demonstrates the applicability of this new paradigm to induce cardiovascular reactivity.
Mean heart rate (HR), spectral power of HR variability in a LF and HF-band, a normalized power ratio, and baroreflex sensitivity (BRS): all these are parameters which somehow reflect sympathetic and/or vagal modulating influence on the SA-node rhythm. In this study the stationarity of these parameters in 5 minute windows is investigated during 2 hours of steady state (sitting rest condition). Mean HR appeared by far the most stable parameter. HR LF-power was the least stable parameter, showing the most variance. Even the normalized power ratio, LF/(LF+ HF), although more stable, appeared to be rather non-stationary. The hypothesis that these results could be explained by a marginally stable baroreflex BP control system, which causes a “waxing and waning” resonance power (LF-power) in HR and BP, was tested by a model of short term BP control. Baroreflex sensitivity, being a system parameter, estimated from spontaneous variations in BP and HR, showed considerable variance in the experimental data too, not reproduced by the model. The rather low coherence found between BP and HR variations makes this BRS estimation technique less reliable, and may at least partly explain this low stationarity of BRS.
The baroreflex constitutes the only hitherto known buffer of rapid arterial blood pressure fluctuations. In order to investigate the influence of sinoaortic and cardiopulmonary baroreflex pathways and nitric oxide (NO) on the dynamic properties of short- term blood pressure control, we determined the power spectra of 24h- blood pressure time series of adult, conscious, freely moving foxhounds under a normal sodium diet. This was done in the intact state (N=6), during blockade of NO-synthesis via a bolus injection of the false substrate NG-nitro-L-arginine ((L-NNA), 16.5±2mg/kg body weight i.v., N=5) and in animals devoid of baroreceptor and cardiopulmonary reflexes (N=5). After L-NNA, blood pressure (BP) increased from 116±4 to 137±6mmHg (P<0.01), heart rate decreased by roughly 30beats/min to 68±3beats/min (P<0.01). The power of spontaneous fluctuations in blood pressure within the frequency range of 0.1Hz - 0.5Hz was tripled by L-NNA (P<0.05). By comparison, total sinoaortic and cardiopulmonary denervation increased power of slower oscillations (<0.1Hz) by a factor of 4.7 (P<0.05), whereas the power within 0.1Hz- 0.5Hz was not significantly different from the control values.
It is concluded that NO and the baroreceptor reflex both play an important role as physiological blood pressure buffers. The NO - system influences predominantly rapid (0.1 - 0.5Hz) and the baroreflex slower fluctuations (<0.1Hz) of arterial blood pressure.
A new statistical method for expressing bonds between 2 variables is proposed, which 1) does not require any hypothesis and 2) may quantify a bond that varies over the ranges of variation of both variables. It allows to quantify the statistical dependence between 2 events with a coefficient, Z, using probabilities of each event and their conditional probability. When applied to blood pressure (BP) and heart rate (HR), the method gave the statistical dependence between 2 discrete values of the variables, expressed as amplitude intervals. Control rats, rats submitted to sinoaortic denervation (SAD) or to an early chronic sympathectomy (SNX) were included in a first study. BP was continuously recorded in baseline conditions during 24 hours in control rats and 30 minutes in SAD and SNX rats. The occurence frequency and the Z value were computed for each couple of systolic BP (SBP) and HR beat-to-beat values. Contour line representations of 3-dimensional Z histograms showed reproducible dependence zones in the (SBP, HR) plane. In control rats, dependence zones were found in the modal class and in 3 zones corresponding to 1) low SBP associated with high HR, 2) high SBP associated with high HR and 3) high SBP associated with low HR. Dependence zones 1 and 3 were not present in SAD rats and dependence zone 2 disappeared in SNX rats. These results suggest the participation of the baroreflex in the dependence between SBP and HR values located in zones 1 and 3 and of the sympathetic nervous system (SNS) in the dependence between SBP and HR values located in zone 2. A second study tested the ability of Z analysis to estimate baroreflex sensitivity (BRS), using mean BP (MAP) and HR values. Dependent (MAP, HR) couples located in zones 1 and 3 were selected and a linear regression was performed. The regression slope, taken as the index of BRS, was strongly correlated to BRS estimated with a pharmacological method. In conclusion, the concept of statistical dependence was applied to BP and HR values taken as statistical events. Z coefficient allowed to separate couples of BP and HR values related to baroreflex activity from those related to SNS activity. In addition, it allowed to estimate in a global manner the BRS from spontaneous BP recordings.
Two methods of assessment of baroreflex sensitivity were compared in eight supine healthy volunteers during repeated baseline measurements and various conditions of cardiac autonomic blockade. The spontaneous baroreflex method involved computer scanning of recordings of continuous finger arterial pressure and electrocardiogram to locate sequences of three or more beats in which pressure spontaneously increased or decreased, with parallel changes in pulse intervals. The mean regression slope of all these sequences during each study condition was considered to represent the mean spontaneous baroreflex slope. In the drug-induced method, sigmoidal curves were constructed from data obtained by bolus injections of phenylephrine and nitroprusside; the tangent taken at the resting pressure of each these curves were compared with the mean spontaneous baroreflex slopes. The two methods yielded slopes that were highly correlated (r=0.96, P<0.001), and significant but similar intra-individual baseline variability. Atropine virtually eliminated the baroreflex slope; subsequent addition of propranolol did not alter it further. Propranolol or clonidine alone increased average baroreflex slope to the extent that they increased resting pulse interval (r=0.69-0.83). The spontaneous baroreflex method provides a reliable, non-invasive assessment of human vagal cardiac baroreflex sensitivity within its physiological operating range.
The sensitivity of the baroreceptor-heart rate reflex has been suggested to have a prognostic value in a number of diseases. In particular, mortality after myocardial infarction, in heart failure patients and in diabetic patients seems to be inversely related to the sensitivity of cardiac baroreflex modulation. These data were obtained, however, by means of traditional laboratory tests, which are affected by important limitations. A deeper insight into the features of daily life baroreflex cardiovascular control can now be obtained by techniques which allow spontaneous baroreflex sensitivity to be assessed in daily life conditions, without need of any intervention on the patient. The various methods currently available to assess spontaneous baroreflex sensitivity are hereby discussed, focusing on the similarities and differences between them and with the traditional laboratory approaches. In particular, the characteristic features and the possible clinical usefulness of the sequence method will be discussed more in details.
Although the so-called low-frequency (around 0.1Hz) components of heart rate variability are an accepted marker of sympathetic activity to the heart and the circulation, their origin remains elusive. In order to characterise the frequency response of the heart and the vessels to the stimulation of the arterial baroreceptors we applied the neck suction technique with a sinusoidal function at different frequencies, and studied the changes in the cardiovascular system. In all subjects the neck suction induced RR interval oscillations at both low- (0.10Hz) and high (0.20Hz) frequencies, but with greater response at 0.1Hz, while blood pressure and skin blood flow oscillations responded only at low frequency (low-pass behaviour). To assess whether this kind of response could have played a role in the genesis of spontaneous LF fluctuations we studied the changes induced by an impulsive neck suction during post-expiratory apnoea: if the LF were due to a resonance in the baroreceptor loop, a single perturbation of the cardiovascular system would have produced oscillations in the cardiovascular system similar to those observed spontaneously. We found that, after the suction, a damped oscillation (whose period was similar and correlated to that of the spontaneous LF) was generated in the cardiovascular system, suggesting that the LF could be indeed generated by a peripheral mechanisms, primarily involving the baroreceptors. In further support of this hypothesis, we also found that counteracting the respiratory fluctuations in blood pressure by appropriately timing a sinusoidal neck suction, we significantly reduced not only the respiratory fluctuations in RR interval, but also the LF. Nevertheless, during undisturbed apnoeas, we observed spontaneous oscillations in the LF range (though slower than those observed during breathing). This observation contrasts with the hypothesis that a resonance in the cardiovascular system is the only source of the LF fluctuations, (since after removing the effects of respiration the LF should disappear as well) and suggests the contemporary presence of other factors, such as a central oscillator, tuned at a similar frequency. In conclusion, the arterial baroreceptors exert a strong, though probably not exclusive influence on the cardiovascular fluctuations, through both sympathetic and vagal activity; the LF appears to be generated not only by a resonance in the cardiovascular system, but also by other independent factors, such as a central oscillator.
We describe a method which makes it is possible to synchronise spontaneous oscillations of skin blood flow to the rhythm of an external thermal oscillatory stimulus. This test can be done totally noninvasively both in adults and newborn infants to study cardiovascular control system in health and disease.