Ebook: Computer Analysis of Cardiovascular Signals
Over the last two decades, the analysis of blood pressure and heart rate variability has been increasingly used to obtain information on the mechanisms responsible for cardiovascular regulation in different physiological and pathological conditions. The evidence gathered so far has strongly supported the ability of this approach to provide unique insights into the factors involved in cardiovascular control. However, studies making use of these techniques have also underlined a number of problems concerning both the selection of the most appropriate analysis procedure to be used in a given situation and the interpretation of the results thus obtained. In this book volume leading experts from both the technological and the medical milieu have deeply addressed these issues. New tools for describing the regulation of the heart and the peripheral circulation in terms of 1/f processes, nonlinear and chaotic and complex systems have also been discussed. The overall result is a collection of contributions which provide the most recent development in this stimulating field.
Computer analysis of blood pressure and heart rate variability is increasingly employed to obtain information on the mechanisms responsible for cardiovascular regulation in different physiological and pathological conditions.
Various aspects related to the methodology and the clinical relevance of this approach were previously addressed in the book “Blood Pressure and Heart Rate Variability”, published by IOS Press in 1992.
Since then, progress has been made in the field of data measurement, data analysis as well as in the physiological and clinical interpretation of the results. For example, new siftware procedures for the estimation of time-frequency distributions have been developed to dynamically evaluate the time modulation of spectral powers in different daily life conditions. Moreover, methods have been implemented for the broad band spectral analysis of blood pressure and heart rate variablity. These techniques resulted in important new tools for the understanding of the regulation of the heart and the peripheral circulation in terms of 1/f processes and non-linear systems. Furthermore, recently obtained experimental data and novel mathematical models have shed new light on the links between blood pressure and heart rate and on the interrelationship between the variability of multiple cardiovascular signals and respiratory activity. Such a multivariate approach to the analysis of variablity in cardiovascular signals has turned out to be much more informative than the traditional univariate methodologies and has been shown to provide a deeper insight into cardiovascular regulation.
The present book includes the contributions on these and other innovative issues given by leading experts in two recent workshops held under the patronage of the European Society for Engineering and Medicine and the European Society of Hypertension.
We hope that investigators working in this field will find in this volume a comprehensive and updated state-of-the art and a helpful reference for their research and clinical activity.
Marco Di Rienzo
The analysis of very rapid fluctuations in heart rate (heart rate variability, HRV) and (systolic) blood pressure (BPV, Blood Pressure Variability) is hampered by the fact that neither the cardiac events, nor the systolic blood pressure values are regularly spaced in time. It is shown that the beat-to-beat variability in systolic blood pressure values is due to both the timing of these values (closely related to the cardiac event series) and an external influence. The well-known model for the SA-node, viz. the Integral Pulse Frequency Modulator (IPFM) appears to operate as an interface between continuously varying external influences and the irregularly spaced values of blood pressure. Three algorithms for the estimation of the spectrum of this external influence are introduced. On the basis of simulation studies it is discussed which method is to be favoured.
The aim of the chapter is to review aspects of the analysis of cardio-respiratory interaction using spectral estimation methods. The first section of the paper provides background on nonlinear physiological oscillation, entrainment phenomena and computer modelling. Entrainment phenomena are described in some detail including: frequency pulling and nonlinear modulation. These phenomena are illustrated using the examples of respiratory sinus arrhythmia and the baro-receptor reflex. ARMA and AR techniques are reviewed and compared to the FFT. Experimental results on the interaction of heart rate, blood pressure and respiration are presented; it was shown that ARMA techniques provided evidence of short term independent physiological mechanisms whereby respiration affects heart rate and blood pressure. In the second section of the chapter time-frequency representations are introduced and the Short-Term Fourier Transform (STET), Wigner–Ville Distribution (WVD) and wavelet transforms are compared on a theoretical basis and applied to heart rate and respiration data, as well as test data, to highlight their differences. The first section of this part of the chapter points out a number of apparently complex phenomena in blood pressure, respiration and heart rate waveforms. These phenomena were shown to be more clearly understood by the application of nonlinear dynamic theory, coupled to computer modelling. In the second section the main advantages of the time-frequency methods were highlighted. These are their robustness and the fact that these methods do not require the data being analyzed to be stationary and as such represent an important development in the beat-to-beat study of cardio-respiratory control.
A comparative evaluation has been made among some methods of power spectrum density estimation, suited to non stationary time series. The evaluation has been performed on artificial and on real data which were obtained during autonomic tests. The study has documented the superiority of the estimation based on time-frequency joint distributions, particularly when a sharp time resolution is required for the analysis of the cardiovascular system variability. The problem of the rejection of cross terms artefacts has been considered as well, by implementing several techniques.
In the present paper the use of wide-band spectral analysis to simultaneously investigate fast and slow components of blood pressure (BP) and heart rate (HR) variability is addressed from both a methodological and a practical point of view. After a short review of the technical problems involved in this kind of analysis, the results obtained by the wide-band spectral approach in evaluating the effects of sino-aortic baroreceptor denervation (cats) and aging (humans) on the different components of BP and HR variability are presented. An important finding from use of wide-band spectral analysis is that the power of the BP spectral components tends to be inversely related to the frequency (1/f trend), namely, the slower the BP fluctuation the higher its contribution to the overall BP variance. The occurrence of slow BP fluctuations of pronounced magnitude, however, may appear in apparent contrast with the anti-oscillatory action of the arterial baroreflex. A possible explanation for this paradoxical phenomenon is that also the sensitivity of the baroreflex is modulated in a 1/f fashion. Experimental data supporting this hypothesis are shown.
We have investigated the cardiovascular dynamics during physiological states such as sleep and wakefulness. Their dynamics are modeled as a multivariate autoregressive time series. Mutual geometric relationships among state-dependent dynamics are obtained by means of the Kullback–Leibler divergence. The dynamics during slow wave sleep is found to be markedly consistent through records and subjects compared with waking states and rapid eye movement sleep. The state-dependency of cardiovascular dynamics and their mutual relationships are suggested to uniquely characterize the autonomic regulation of the cardiovascular system.
The biological rhythm is basically subject to 1/f fluctuations. The typical examples are heart rate and blood pressure. The present author has found that, in addition to the heart rate, spontaneous discharge interval of a giant neuron of African snail has 1/f fluctuations and hand clapping intervals also have 1/f fluctuations. Furthermore, the axon as a transmission line of biological signal has a strong nonlinearity, which modulates the time relation of launched action potential impulses; the time relation of randomly launched impulses is modulated by this nonlinearity and eventually it reached 1/f fluctuations. The biological body has a special affinity to 1/f fluctuations. Computer simulation shows that 1/f interval fluctuations are deeply related to 1/f fluctuations of membrane admittance to ion currents. The physical origin will be the same as conductance fluctuations of electric conductors which have been widely observed in any type of electric conductor. Finally, a possibility is discussed that 1/f fluctuations in biological rhythm and signal transmission have special meaning as regards stable parallel control.
Multi-variate identification methods applied for the analysis of heart period (HP), systolic arterial pressure (SAP) and respiration beat-by-beat series are discussed. A method for multi-variate spectral decomposition and for a classification of rhythm sources is presented and applied in a preliminary study on conscious dogs during different manoeuvres of sympathetic stimulation and enhancement of low frequency (LF) waves. Different contributions to LF oscillations are detected at control and different activation modes are found depending on the kind of the stimulus: overall increase of disturbances during nitroglycerine infusion, increased rhythmicity of sinus node modulation during coronary artery occlusion and increased resonance of baroregulation during bilateral carotid occlusion. In conclusion, multi-variate spectral decomposition provides a classification of variability rhythms more complete than that provided by mono-variate methods, which is based only on the position on the frequency axis.
That blood pressure is highly variable over 24-hour, yet would be tightly controlled by the baroreflex system is a baroreflex paradox. Not only blood pressure varies but baroreflex sensitivity varies with it. Changes in baroreflex sensitivity can cause great changes in blood pressure, as we have shown in model studies. Consequently, the baromodulation hypothesis says that sensitivity modulation is the primary mechanism in the baroreflex that changes arterial pressure. The baroreflex then becomes not just a pressure stabilizer but a pressure controller: in interaction with other inputs such as coming from higher brain centers, it moves the blood pressure from one desired level to another. This system is effective, fast, and safe. We prefer not to use the word resetting for this mechanism but to reserve ‘resetting’ traditionally for changes in the receptor function. For resetting the baroreflex as a functional unit the word ‘baromodulation’ could be used. From recently recorded 24-hour 1/f blood pressure spectra it has again been confirmed that speaking of a baroreflex setpoint, even of a setpoint that is reset occasionally, is less appropriate. Blood pressure changes continuously and intentionally. Baromodulation is capable of doing that.
In order to study respiratory modulating effects on blood pressure and heart rate over a wide frequency range and in different autonomic states, we developed a computer model of the baroreflex blood pressure control system. The model combines simple beat-to-beat hemodynamics with detailed dynamic neural control of heart rate and vasomotor activity. The model is capable of simulating different heart rate and blood pressure steady-state values as a result of a changing sympathovagal balance, and accurately simulates beat-to-beat variability due to respiration and 0.1 Hz resonance. The model is a powerful instrument in formulating and testing hypothesis concerning the origin of RSA, frequency dependency of RSA, and short term blood pressure modulations. The main characteristics of the model are discussed, as well as its use in the interpretation of two experimental studies, standing versus supine posture, and sympathetic (spinal) lesions.
A computer model for the circulation was developed that takes gravitational effects into account. Gravity was supposed to act on the filling of the left and right atrium and the hydrostatic distance between heart and baroreceptors. The model is based on difference equations describing essential conditions of the circulation, like heart period, systolic pressure or stroke volume, on a beat-to-beat basis. It was possible in this model to simulate the cardiovascular dynamics after a passive tilt from upright to supine, assuming only baroreflex control. However, the dynamic behaviour after a tilt from supine to upright was only wellsimulated when low-pressure baroreceptors (in the atria) were incorporated in the model as well. Depending on the parameter settings of the model sustained 0.1 Hz oscillations in heart rate and blood pressure showed up in the upright situation, due to non-linearities in the model and sympathetic feedback reaction times.
In normal subjects the respiratory sinus arrhythmia (RSA) is considered an index of efferent vagal activity to the heart. The human transplanted heart is thought to be denervated and hence any fluctuation present in the RR variability can be either due to reinnervation or to represent the effect of a non autonomic activity, such as a direct effect of respiration on atrial stretch. We evaluated the presence of spontaneous fluctuations in RR interval by simultaneous power spectrum analysis of RR interval and respiration variability, using uni- and bivariate algorithms.
A small but clear ‘RSA’ was present in all transplant subjects, evidenced by a small fluctuation on the RR interval power spectrum, synchronous and coherent with the main component of the respiratory spectrum. Although the amplitude of this ‘RSA’ was in the range of 1-4% of that of controls its possible artifactual origin could be ruled out by absence of major changes in QRS morphology with respiration. No changes were observed during manoeuvres altering the sympatho-vagal balance (such as tilting), or known to produce a reflex brady- or tachycardia (neck suction or amyl nitrite inhalation) in subjects transplanted 3-20 months earlier. The only condition which modified the RSA in transplants was the increase in ventilation during physical exercise regardless of changes in sympathetic tone. This “RSA’ did not increase with time since transplantation (3-52 months).
These data suggest a non reflex mechanical effect on atrial stretch, from inspiratory increased venous return. In particular cases we have found that the respiratory effect extends also to the so-called low-frequency band, attributed to the sympathetic influence: the simultaneous acquisition and analysis of the respiratory signal is therefore essential for a correct interpretation of the spectral bands, even when the breathing is controlled by a metronome. However, in subjects who had their transplant for more than 20 months we often found low-frequency oscillations (LF) of non-respiratory origin; only in these subjects could amyl nitrite produce an immediate increase in heart rate. In the absence of other known alternative explanations, these LF should, unlike the RSA, be regarded as a sign of functional reinnervation in heart transplanted subjects.
Heart rate variability in the 0.1 Hz frequency range has been reported to provide a non-invasive measure of sympathetic activity, however the technique has not been validated for the quantification of differences in sympathetic activity between individuals. We compared 0.1 Hz power with cardiac noradrenaline spillover and muscle sympathetic nerve activity in conditions of sympathetic denervation, including pure autonomic failure and after cardiac transplantation, and in the context of sympathetic nervous activation, in cardiac failure and in relation to ageing and during application of mental stress and isometric exercise. While there was agreement between techniques in pure autonomic failure and early after transplantation, disparity between spectral analysis and the more direct indices of sympathetic function were obtained for cardiac failure and ageing. Among healthy subjects, in whom measurements were made simultaneously, there was no significant correlation between 0.1 Hz power in either absolute units or normalised for total power and the rate of noradrenaline spillover from the heart to plasma. Heart rate variability clearly provides a functional measure of cardiac sympathetic mechanisms which is dependent on post-synaptic aspects of neural function, in addition to rates of sympathetic nerve firing. One particular application of heart rate spectral analysis in neurophysiological research might be in combination with the cardiac noradrenaline spillover technique, allowing more comprehensive assessment of both neuronal and post-synaptic aspects of the cardiac neuroeffector response.
The cardiovascular system oscillates around several frequencies that are specific for a given domain or territory. From membrane potential oscillations generating action potentials in the myocardium to the blood pressure and the heart rate variability, different mechanisms create rhythmic patterns whose significance is presently extensively investigated.
This study describes the vasomotion in the radial artery of healthy volunteers. The combined long tern measurements of blood pressure and internal diameter by an A-mode ultrasonic echo-tracking device allow us to assess the mechanical properties of the vascular wall. Under resting conditions, the radial artery diameter of the volunteers oscillated with an amplitude 3 to 4 times greater than the pulsatile diameter changes and with a period ranging from 40 to 70 seconds. Decomposition of the heart rate, blood pressure and diameter variability by power spectral analysis techniques showed a very low frequency (VLF) mode ≤ 0.02 Hz for diameter signals that could not be identified for the other two signals in the time domain studied (13 min). The vasomotion of the radial artery induces a significant modification of the mechanical properties of the vessel wall as assessed by cross-sectional compliance- or distensibility-pressure curves.
Although the spectral analysis does not support the influence of a centrally mediated mechanism of vasomotion in the muscular artery of the forearm, further studies will be necessary to verify this hypothesis.
In this study we tested the hypothesis that shear-stress-dependent endothelium-derived nitric oxide (EDNO) release acts as a physiological blood pressure buffer. EDNO was blocked by the false substrate for EDNO-synthesis (NG-nitro-L-arginine (L-NNA), 16.5 ± 2 mg/kg body weight i.v.) over 24 hours. Blood pressure and heart rate (HR) were examined over this period in freely moving conscious dogs (N = 6). After L-NNA, mean arterial pressure (MAP) increased from 116±5 mmHg to 134±5 mmHg (P < 0.01), heart rate decreased from 97 ± 6 to 68 ± 3 beats/min (P < 0.01). The power spectrum revealed a 2.1 fold increase of power in the frequency range between 0.01-0.5 Hz (P < 0.05) in response to L-NNA. These data suggest, that blockade of EDNO provides sustained hypertension over a 24 hour recordings and that EDNO is important as a physiological blood pressure buffer in the frequency range below 0.5 Hz. Thus, the endothelium participates in the control blood pressure variability together with arterial baroreceptors
Spectral analysis was recently chosen to characterize the fast oscillations depending on the activity of the sympathetic and parasympathetic nervous systems. Renin stimuli could impinge on different, i.e. low frequency domain since the time lag to renin-angiotensin system activation is significantly larger. This experimental study was designed to analyze low frequency components of short-term variability of blood pressure (BP) of conscious rats in two conditions where renin was activated. First, a combined blockade of autonomic nervous system and circulating vasopressin was achieved with i.v. chlorisondamine plus a selective vascular V1 receptor antagonist. This treatment determined a 4.2-fold increase in plasma renin activity. Spectral powers of the systolic and diastolic BP and heart rate (HR) were computed in the high (respiratory), mid (0.2-0.6 Hz, MF, Mayer waves mainly of sympathetic origin) and low (0.02-0.2 Hz, LF) frequency bands, as detected by the Fast Fourier Transform technique on consecutive 102 s stationary periods. The blockade resulted in a significant decrease in the MF (sympathetic) power of BP (−53%, P < 0.001) and HR (−69%, P < 0.001). A significant increase in the LF component of BP variability was observed (+43%, P < 0.001). ACE inhibition (enalaprilate i.v.) abolished this component. Secondly, we studied a model of renovascular hypertension (two-kidney Goldblatt hypertension) after 3 weeks and we also observed a marked LF component of BP variability when hypertensive rats were compared to sham operated animals (+73%). Interestingly, losartan i.v. markedly reduced this LF component. Losartan also induced an increase in the MF oscillations of BP and HR in the two groups. A reflex sympathetic activation secondary to All antagonism could contribute to the observed effects on MF oscillations.
Thus, an increase in the LF component of BP variability was observed when sympathetic activity was reduced with autonomic blockade and in a model of renin-dependent hypertension. The dependence of these 20 sec period oscillations upon the renin-angiotensin system activity was demonstrated with pharmacological blockade with enalapril or losartan.
Computer analysis of blood pressure and heart rate fluctuations and the quantification of their interaction in the time (sequence method) or in the frequency domain (spectral analysis) offer a dynamic evaluation of the sensitivity of baroreflex control of the heart (BRS) in daily life conditions. Use of these techniques in the analysis of 24 hour intra-arterial blood pressure and heart rate recordings has shown that in real life BRS is continuously modulated over time and is characterized by a pronounced increase at night as compared to the daytime. In a number of pathophysiological states (hypertension, aging and autonomic failure), average 24 hour BRS is markedly lower than in normal subjects and its day-night modulation is blunted to a variable extent. These time domain and frequency domain techniques provide generally superimposable results and represent unique tools not only to quantify differences in BRS between different steady-state conditions, but also to track fast changes in BRS induced by behavioural stimuli on a minute-to- minute basis. They offer a deeper insight into daily life cardiovascular regulation in normal and diseased conditions with no need of external interventions.
In this chapter the effects of mental effort during task performance are studied on heart rate, blood pressure and its spectral variability measures as well as effects on baroreflex sensitivity also using a spectral method. In particular, differential effects of task complexity is looked at. On heart rate about the same effects are found as in other laboratory studies on mental workload: an increase of heart rate and a decrease in variability. The pattern in blood pressure is about the same: an increase in pressure during mental effort and a decrease of blood pressure variability. In most cases also baroreflex sensitivity is decreased.
The second part of the paper is directed to differential effects between tasks. It is found that the mid and higher frequency part of the heart rate spectrum shows the largest effects of task complexity. Changes in respiratory pattern during task performance are studied to get a better insight in the possible reasons for these differences. It is concluded that differences in breathing pattern can possibly partly explain rest-task changes in variability measures. However, differences in spectral measures between tasks can not be attributed to respiratory patterns. These latter effects have to be ascribed to an additional vagal inhibition during complex task performance.
The relationship between sympathetic activity and specific spectral components of blood pressure (BP) variability has been investigated in conscious Wistar Kyoto (WKY) rats and in spontaneously hypertensive rats (SHR), the latter being characterized by sympathetic overactivity.
The study was based on the evaluation of BP spectral powers in SHR and WKY rats with intact sympathetic nervous system and after abolition of efferent sympathetic activity obtained by chemical sympathectomy. In each spontaneously behaving rat, systolic blood pressure, diastolic blood pressure and pulse interval (the reciprocal of heart rate) were monitored beat-to-beat for 90 min. The spectral powers of these parameters were calculated by FFT technique and integrated over three frequency bands (HF 0.8-3 Hz, MF 0.1-0.6 Hz and LF 0.025-0.1 Hz). The obtained data showed a non-univocal relationship between sympathetic activity and the MF and LF components of BP variability. Furthermore, to estimate the residual components of BP variability in sympathectomized SHR and WKY rats also beyond the LF band, a wideband analysis from 0.0007 to 3 Hz was computed. Our results indicated that the abolition of sympathetic efferent activity did not cancel the differences of BP spectral profiles between SHR and WKY rats, thus suggesting that these differences are due, at least in part, to structural and/or humoral factors.
The mechanisms underlying arterial blood pressure (AP) and heart rate (HR) beat-to-beat variability were investigated using spectral analysis in conscious genetically normotensive (LN) adult rats from the Lyon strain. Basal AP and HR spectra exhibited peaks in low- (LF: 0.27-0.74 Hz) and high- (HF: 0.75-3.85 Hz) frequencies. The LF oscillations of systolic AP, and even more of diastolic AP, could be attributed to the sympathetic nervous system influence as, after destruction of the peripheral sympathetic nerves, the LF peak disappeared and the LF power spectral density (PSD) was highly reduced. Ganglionic blockade with chlorisondamine combined with a restoration of the basal mean arterial pressure (MAP) level decreased LF PSD of MAP in control rats in a larger extent than sympathectomy. No relationship was found between the MAP response to chlorisondamine, taken as an index of the sympathetic vasomotor tone, and the basal LF PSD. The role of the baroreflex in the spectral characteristics of MAP and HR was also investigated. In rats with a chronic sinoaortic baroreceptor denervation (SAD), PSD of MAP was reduced in the LF band. Transfer function analysis between MAP and HR showed that, in control rats, coherence was high for frequencies surrounding the LF and HF peaks. In SAD rats, coherence in the LF band was abolished but maintained in the HF band. In conclusion, nearly 80% of the LF PSD of AP depend upon the autonomic nervous system activity and the activation of the peripheral sympathetic nerves contributes mainly to the production of LF spectral power. The baroreflex accounts for half this power and for the coherence between MAP and HR oscillations in the LF band only.
Autonomic abnormalities are an important component of the pathophysiology of patients with cardiac failure. Power spectral analysis of heart rate variability gives a noninvasive means to investigate this abnormality and a prognostic tool in the management of this syndrome. Further studies are, however, necessary to establish the reproducibility and physiological significance of the parameters estimated by this technique.
In 16 patients two weeks after the first uncomplicated myocardial infarction and in 10 control subjects we analyzed heart rate and systolic arterial pressure variabilities. During resting controlled conditions, patients presented an increased LF component of RR and systolic arterial pressure variabilities and a diminished HF component of RR variability. These changes were consistent with a sympathetic excitation and with a reduced vagal tone. The gain of baroreceptive mechanisms was assessed by analyzing the relationship between the spontaneous oscillations present in RR and systolic arterial pressure variabilities in the LF and HF frequency range and was found to be smaller in post-myocardial infarction patients than in control subjects. This non-invasive approach seems therefore capable of providing important information on the alteration on neural mechanisms controlling heart period and vasomotion after myocardial infarction.
This manuscript attempts to link the physiological uses of measuring short term heart rate variations with the finding that quantification of long term heart rate variability is predictive of sudden death after myocardial infarction. The mechanisms which underlie short term variations of heart rate and the data which have provided the association between long term variations of heart rate and sudden death are reviewed. An argument is then made to support the hypothesis that changes in long term heart rate variability may be driven more by activity level and the heart’s response to that activity than by abnormalities of sympathetic and parasympathetic heart rate control.