CA1298656C - Method and apparatus for the assessment of autonomic response by broad-band excitation - Google Patents

Abstract

Abstract of the Disclosure This invention provides a rapid, noninvasive technique for quantifying the dynamic response of the autonomic nervous system (ANS) to perturbations it senses over a broad range of physiologically relevant frequencies. The technique involves two steps. First, a physiologic parameter sensed by the ANS is subjected to a broad-band perturbation as an input signal while a physiologic parameter modulated by the ANS is monitored as an output signal. Then, the transfer relation between input signal and output signal is determined.
The computed transfer relation is then readily interpretable in terms of responsiveness of the various limbs of the ANS.

Description

~2~ S~

~ 1 --!s'.. -, a407w - METHOD ~ND APPARATUS FOR THE ASSESSMENT OF
AUTONOMIC RESPONSE BY BROAD-~ND_EXCITAq'ION

Backqround of the Invention :-- Ths GovQrnment has rights in this invention ~ ",,-pursuan~ to Grant Number NAG2-327 awarded by the National Aeronautics and Space Administration~
Tha role of the cardiorespiratory system i~ to maintain perfuslon of appropriately oxygenated blood to the variou~ ~issues and organ~ of the body. This operation i~ carefully regulated by the ANS, which continuously senses hemodynamic variables that reflect .^ the ad~quacy of tissue perfusion, such as arterial blood pressure and oxygen contenk, and then efects changes in resplration, cardlac output, and vascular resistance so 7i~ 15 as to maintain these variables within a narrow range.
In this regard the ANS, combined with the effector ~ organs it regulates, serves as a feedback and control .. system. It i8 well known that the performance of this ;l~ feedback system becomes compromised in a variety of pathological conditions, such as heart failure, hypertension, shock, and diabetes, to name but a few.
~; ~uantitative assessment of the performance of the ANS
would tharefora be a vikal tool in a wide range of ~-; clinical situations. Until now, however, there has . 25 existed no reliable noninva~ive technique for achi~ving : this quantitaive a~sessment of autonomic function.
Performance of a feedback system can be ascQr~ained by measuring the degree to which it can effect the appropriate changes in the variables it controls in r~sponse to fluctuations in the variable~ it .~.
A ~ ~

;'~`
,ii ~2~a6s~

: - 2 -senses. Thorough evaluation of the system's performance - raquires testing its response to a set of perturbations that fully represents the range of natural perturbations : the system experiences. Most of the previously described technigues developed to characerize the ANS
involve either not per~urbing the system at all and measuring spontaneous fluctuations in hemodynamic variables ~1, 2, 4, 5, 6, 7, 9, 10, 14, 16, 19, 25], or mea~uring the system's response to perturbation~ applied at only a si~gle freguency a~ a time [3, 8, 11, 12, 13, 17, 20, 21, 22, 23]. The numbers in brackets refer to refQrences listQd in Appendix A. In the first case, the investigator'~ ability to examine system behavior is - limited to the generally narrow frequency bands within which spontaneous fluctuations perturb the system.
Furthermore, as thQse spontaneous perturbations are o~ten in the form of an unmeasurable noise component, ! one is left with the task of attempting to infer in~ormation about the system from measurements of its output signals only. In the second case, since .~ information about the system can be acquired at only one frequency at a time, a complete characterization using this sort o approach necessitates repeating the test procedur~ at many different frequencies. Such a ~- 25 protocol is cumbersome, as it requires a long time to ~`, complete. In addition, any system characterization derived from data pooled from multiple test trials may Y~ possess artifacts related to changes in test conditions ; from OnQ trial to another.
. 30 ~ technique that would allow for rapid :- simultaneous acquisition of all information required to ~`~ fully characterize the dynamic behavior of the ~NS would therafore represent a tremendous improvement in the pra~tical utility of this clinical tool. The approach r~

I ~.,--;

12~86S6 we have dQveloped requires a slngle typically six-minute :. period of data collection from which th~ system characterization over all frequencies of interest is derived.
In order to obtain a charac~erization of system responsivenes6 over the entire range of physiologically impor~ant requencies (generally considered 0.0 -0.5 Hz) from a single six-minute record of input and - output signals, the input, or perturbing, si~nal must simultaneously contain components at all frequencies ~~ within this range. The input signal must therQfore be of a "broad-band" or "white noise" nature.
The use of a broad-band excitation in exploring the machanical proparties of cardiovascular system was ;~ 15 described by Taylor [24]. He applied randomly timed pacing impulse~ to the cardiac atria of experimental '~ animals via surgically implanted pacing wires, and charactQrized the transmission line prop~rties of the -.~ arterial tree through the computation of an impedance function in the frequency domain. He found that the effective impedance of the arterial system was depressed ; at frequencies less than 0.Q3 Hz. Taylor pointed out ~` that this behavior indirectly demonstrates the presence . of compensatory mechanisms within the ANS that modulate arterial resistanca at low frequencies. Ha made no i~?; attempt, however, to explicitly characerize the dynamic respons~ of the ANS in performing this function. In ; contrast, the technique we present here provides a means for probing tha specific transfer properties of the ANS, entirely noninvasively. Furthermore, we have found that our t~chnique is surprisingly sensitive in its ability `'. to quantify subtla change~ in the performanco of this -~ feedback system, '''", :;
..5 , .
,~t?
'Yt ~ .

2915 E;S~i 4 737~6-9 Summarv of the Inventlon The present invention assesses response of the autonomic nervous system by ~ubjecting the autonomic nervous system to a broad-band perturbation. A hemodynamic parameter modulated by the autonomic nervous system, such as heart rate, is monitored and the transfer relatlon of the broad-band perturbation and the hemodynamic parameter is computed. A suitable perturbation of ~he autonomic nervous system is a broad-band perturbation on respiratory activity as measured by instantaneous lung volume.
Other suitable perturbations of the autonomic nervous system will be described hereinbelow.
According to a broad aspect of the invention there is provided a method for assessing autonomic response comprising:
applying a broad-band perturbation to a physiologic input signal sensed by the autonomic nervous system;
monitoring a physiologic output signal modulated by the autonomic nervou~ system; and computing the transfex relation between the physiologic input and output signals.
According to another broad aspect of the invention there is provided a method for assessing autonomic response comprising:
applying a broad-band perturbation to a physiologic parameter - which is the input signal sensed hy the autonomic ner~ous system;
monitoring an output signal which is a physiologic parameter which is modulated by the autonomic nervous system; and computing the relatlonship between the input and output signals.

.~ ~

- ~9136~
.
4a 73766-9 According to another broad aspect of the invention ~here is provided apparatus for assessing autonomic response comprising:
means for subjec~iny a physiologic parameter sensed by the autonomic system as an input signal to a broad-band perturbation;
means for monitoring an output signal which is a physiologic parameter modulated hy the autonomic nervous system; and means for co~puting the relationship between the input and output signals.
~rief Description of the Drawin~
The invention disclosed herein will be understood better with reference to the drawing in which:
Figure la is a graph of a respiration spectrum;
Figure lb is a graph of a heart rate spectrum;
Figure 2a i5 a graph of the transfer function magnitude of the data in Figures la and lb;
Figure 2b is a graph of the transfer function phase of the ~ata of Figures la and lb;
Figure 2c is a graph of the coherence computed from the data of Figures 2a and 2b;
Figures 3a, b, and c are graph~ similar to those of Figure 2 i~ which the subject is æupine; and Figure 4 is a schematic illustration of a hardware implementation of the present invention.
Description of the Preferred Embodiment~
Our techni~ue requires first identifying appropriate hemodynamic or respiratory variables to represent input and output si~nals for the ANS. An appropriate input signal is one that reflects cardiorespiratory activity sensed by the ANS, and is . ~ . .

.. controllable to allow imposition of 3 desired waveform G., on it. The output signal must be easily and ~`- noninvasivaly measurable and should reflect hemodynamic ;: changes effected by the ANS.
While the ANS per~orms its function through modulation of multiple hemodynamic variables, heart rate i6 a particularly use~ul metric of autonomic function.
This i~ bccause in the absence of autonomic control (e.g., after ablation of the neural networks through which the ANS communicates with its effector organs), the heart rate remains virtually constant. Heart rate is thus a relatively pure reflection o autonomic i~ctivity, and $or this reason sQrves as the exemplary output variable in the technique described here.
Since the ANS senses a wide variety of cardiorespiratory variables, a ranga of suitable ~hoices are available for use as the input signal. In the examplQ desaribed below, for instance, we employ -~ instantaneous lung volume as the input signal and impose broad-band perturbations on respiratory activity. Other ~` convenient choices for the input variable are discussed in the section on generalization of the method.
~` As noted above, in order to afford a complete characterization of the ANS from data acguired during a ~ingle si~-minute test trial, the waveform imposed on ..~, ~, the input 3ignal must contain components at all frequQncies of interest simultaneously. Any sufficiQntly broad-band waveform will serve to excite the ANS with multiplQ simultaneous frequency components. The choice of a particular waveform is made on the basis of what is conveniently realizable ~or imposition on the input variable used, and what is most suited to the analysis technique employed for determination of the transfer relation between input and ,., , A~
,;.
..
_ ~L298~i~6 - : - 6 -,. .
~; output signals. In the example section, we describe one type of broad-band waveform that is particularly easy to impose on the instantaneous lung volume signal.
Once input and output signals have been : 5 recorded during the test period of broad-band excitation, any of a wide variety of signal processing techniques may be utilized to compute the transfer relation between the signals. These inalude both time , and frequQncy domain approaches, and within either catagory are techniques that make various a priori ; assumptions about the system behavior. The choice of ~- which ~ignal processing method to use depQnds on what attributo of the sys~em' 8 behavior is to be probed. In the example presented below, WQ use a frequency domain 15 approach that yields transfer ~unction magnitude and phase plots derived from the cross-spectrum of the input and output ~ignal~.
Example In this example we demonstrate the utility of 20 our technique by showing how it can be used to ; . .;
'"',5,'' characterize the behavior of components of the ANS
. involved in mediation of heart rate fluctuations resulting rom respiratory activity. Both sympathetic and parasympathetic diversions of ths ANS contribute to 25 this behavior. Since many of the components involved in i this control path ( 2 . g ., many of the brainstem centers, ^;~ the heart'~ pacemaker, and the neural pathways through / which the ANS effccts changes in cardiovascular function) are also involved in other hemodynamic reflex m 30 arcs, the transfer relations found in this exampla have broad implications regarding general ANS behavior.
The input signal we used, which we denote x(t), was the in~tantaneous lung volume, measured using a noninvasive volumetric transducer (Respitrace). The .:

. . .
.. .
~j I

~a~9~3~s6 output signal y~t3 was instantaneous heart rate, derived -~ from the elec~rocardiogram (ECG) using an event tachometer algorithm. We characterize ~he response - characteristics of the ANS with a complex transfer function Sxy(f) H(f) = (1 wherQ SXx~f3 is the power spectrum of x(t) and 9xy(f) is the cross-spectrum between x(t) and y(t).
The complex transfer function H(f) is then decomposed into magnituds and phase components, ¦H(f)¦ and ~(f~ respectively, from the real part HR(f) and the imaginary part HI~f) of the complex transfer function as follows:
- ¦ H(f~l a ~(HR(f))2 + (HI(f))2] / ~2a~

, ., H I ( f ) ~(f) - tan~l - (2b) ~ . HR(f~
, ` The transfer function magnitude ¦H(f)¦ reflect~ the ;~ degre~ to which input signal content at a frequency f becomes manifest in output signal content at the same reguency. Th~ phase ~f) indicates what fraction of the period corresponding to the frequency f the output - ~ignal is delayed with respect to the input. This particular repr~sentation of the transer characteristics i8 useful because it facilitates the development of equivalent electrical circuits to model tha sys~em behavior.
~` As discussed above, th~ choice of respiratory activity as the input signal to the ANS is not unique;
-~' , . ~ j ~ ..
--_ .

~2g~s~

other variables may be chosen which can similarly be ~;. measured and perturbed so as to provide the necessary - input excitation. Respiration is particularly useful as the input signal si~ce in addition to being easily and noninvasively controllable, it is known to in~luence autonomic activity through several physiologic mechanisms. First, the cycle o inspiration followed by expiration causes fluctuations in intrathoracic pressura ; which modulates arterial pressuro both directly and through modulation of cardiac filling. Arterial pressure fluctuations are then sensed by the ANS vla carotid and aortic baroreceptors. Second, during inspiration, str~tch receptors in the chest wall are activated and send signals to the brainstem which affect ANS activity. Third, and perhaps most important, there is likely a direct neural link between the respiratory ','~,~7 control contQr in the brainstem and other ANS centers such as those that control heart rate.
An important aspect of the implementation of }: 20 our techniquQ in this example is the way in which the :~; perturbing signal, in this case the instantaneous lung volume, is made broad-band without significantly altering the normal respiratory mechanics. In particular, our method preserves the subject's ~ormal residual (post-expiratory) lung volume and allows him to . titrate the depth of his inspirations so as ko maintain normal blood gases. He is instructed to initiat~ a~
-`~ inspiratory/expiratory cycle each time he is cued by an .~ audible tone. The tones are generated by a computer, programm~d to space the tones evenly in time at some preset rate for a few minutes so that the subject can --~ find a comfortable depth of inspiration. The program then changes modes so that the tones occur with irregular lntervals for the next six minu~es, but a~ the .~.

~2~6~6 g same mean rate as during the constant interval sequence. The program can be easily modified to use any desired distribution of intervals between successive :- tonQs. Of course, cues other than audible cues may be used. An example is visibl~ cues.
The resulting in~tantaneous lung volu~e signal x(t), as measursd by impedance plethysmography, approximates the result of convolution between a -j sequence of unevenly spaced delta unctions q~t) and the mean ~ingle-cycle respiratory waveform r(t). Thus, x(t) - q~t) * r(t) (3) Where "*" is the convolution operator. This leads to the following relationship in the frequency domain:
Sxx(f) - Sqq(~ R~f)12 whera S~x(f) i6 the power spectrum of x(t), Sgq(f) is the powar spectrum of the pulse sequence q(t), and IR~f~l2 is the Fourier transform magnitude squared -~ of the waveform r(t). The input signal x~t) will be sufficiently broad-band if its power spectrum SXx(f) `~ 20 is significantly non-zero for all frequencies f of ~:~ interest. This requires that both S~q(f~ and ¦R~f)¦ are non-zero over the same frequency range. ¦R~f)¦2 falls to zero beyond some .~ frequencyl but i8 assumed to remain significantly non-zero to at least ON5 HzN n any case r(t), and thus ¦R(f~¦2, cannot be modified without altering the subjec~'s respiratory mechanics. The shape of SXx~f) thu~ depends strongly on the nature of Sqq(f) within the frequency band of interest, which in turn depends on - 30 the distribution of intervals used.
. A sensible choice for the distribution of intervals is that of a Pois~on process, since the power ;~. spectrum o~ a sequence of Poisson impulses is a constant ~` over all ~raquencias. The intQrpulse interval ~ , .

65~
. .
~: -- 10 --.
~` distribution pt~t) in this case is a decaying ~-: exponential in t. Thus, pt(t) = ~ e~~ (5) where ~ is the mean occurrence rate of the tones. The difficulty with this distribution is that not only can arbitrarily short intervals occur, but such intervals are in fact favored to occur. In practice, we find a subject has diffiaulty initlating a new respiratory cycle if he is in the midst of an inspiratory/
expiratory cycle when he hears the next tone, even though we instruct him to attempt to do so. For this reason, we modify the distribution so as to prohibit intervals shorter than some minimum interval tmin:
pt(t) = ~e ~ tmin) U(t~tmin) (6) where U(~ equals unity for ~ 0 and zero otherwise.
~-~- With little difficulty, this distribution can be further modified to allow for intervals between both ~ a minimum and maximum limit:
".5S' pt~t)=~e-~tt-tmin)-u(t-tmin)-u(tmax t) ~' is a constant greater than ~ such that ~,~" tmax .: ~f, ~I` Pt(t)dt = 1 (~
tmin - We find that the imposition of such limits on the intervals greatly improves the ability of the subject to ,j$ follow the desired breath generating sequence, while ~- 25 only slightly compromising the broad-band nature of the impulse train q(t). It is important to note that even if the power density of the respiratory signal is not constant, the ability to compute an accurate transfer function H(f) is not diminished since variations in SXx(f) will be compensated for ~hen the quotient : `
. .
: . ~

~ _ .. , ~L29865~

z indicated in equation 1 is computed. However, confidence limits on the value computed for H(f) are expected to be wider in regions along the frequency axis where the input signal power density falls appreciably.
The ordinary coherence function ~xy~f), computed from tha input signal spectrum SXx~f), the output spectrum Sw~f~, and the eross-spectrum Sxy( f ), thQ
-. output spectrum Syy(f)~ and the cross-spectrum Sxy(f) as ollows: ISxy(~) ~2 (f) SXx(~)S~ (f) may be used to provido a quantitative assessment of the confidence, as a function o frequency, of th~ value computed ~or H(f) at each frequency. The important iSSUQ i8 that SXx(f~ maintain a significantly non-zero ' 15 1QVe1 throughout the frequency band of interest, even if the level is not constant.
~` Fig. la shows the power spectrum of the `~; respiratory signal measured for one subject. The distribution of intervals used was pt(t) of equation 7 ~,;' 20 with tmi~ ~ 1 sec, tmaX = 15 sec, and mean interval length of 5 see. Note that while the power density ~-' varies eonsiderably over tha freguency band from 0.0 to ~c~
0.5 Hz, there is nonetheless at least some power at all frequeneies.
The technique described above may be used ~o ~- assess shlfts in the relativQ balanca between sympathetie and parasympathetic activity in tho mediation of respiratory induced fluctuations in heart rate, from ona point in time to another. This i8 illustrated in the accompanying figures. The respiratory and heart rata power spectra shown in .:.
~ r;

~ .

~L~986S6 : Figs. la and lb respectivQly were computed from signals --- recorded whilQ the subject was standing upright during the six-mlnute test period. Note that the heart rate ~ignal contains almost no power beyond 0.2 Hz, dQspite the presence of power in thQ input excitation throughout the frequency band displayed.
Figs. 2a and 2b show the transfer function magnitude and phase plots, where the respiratory and ,,",!'., heart rate signals of Flg. 1 are taken to be the input and output signal6 respectively. The magnitude plot ~Fig. 2a~ demonstrates the prasence of accentuated response regions within the frequency band displayed, most notably around 0.1 Hz. Then from 0.15 Hz to 0.30 Hz, the re~ponSQ magnitude falls gradually, much like a low-order low-pass filter. The phase plot (Fig. 2b) shows roughly linear phase behavior in the region from 0.02 Hz to at least 0.20 Hz, suggesting the presencs of ; a delay element in the control path involved in ; regulating heart rate. (The phase is plotted modulo .-~ 20 2~, so one must visually "unwrap" the plot to ~ appreciate the linear decline in phase as a function of ;~ frequency.) That the autonomic control system regulating heart rate behaves as a low pass filter with a delay i8 highly suggestive of a predominanc~ of sympathatic activity in the control process. Fig. 2c shows the ordinary cohQrence function computed from `l~ these data. A coherence value of unit at a particular frequency would imply perfectly linear operation of the control network and an absence of corrupting noisa in ~he system. Where the coherence func~ion ~alls substantially from unity either the system functions 68 idQally or ther3 exists additive noisQ that makes the transfar function estimate at those frequancies les8 reliable. The coherence function thus serves as a check '';; ~' . ,~, -.
` r .~ ,.
. r ~' .~

12986~;6 ~- on the transfer function computation.
-~ Fig3. 3a, 3b, and 3c are analogous to Fig. 2a, ~b, and 2c except they were computed from data collected ~ while the subject was supine instead of upright. Note ~hat in contrast to the situation when the subject was upright, th~ transfer function magnitude in the supine case remains relatively steady over a broad frequency band from 0.02 to at least 0.30 Hz. Also, in this case, : the phase plot hovers around zero throughout the frequency band displayed, indicating a lack of any significant delay in the control path. The broad-band :- rQsponSiVeneSs of th0 control system without the - presence o~ significant delay in this case ~uggests parasympath~tic activlty predominates in tha mediation o~ respiratory-induced heart rate fluctuations when the subject is supine.
; This shift in autonomic balance from sympathetic predominance to mostly parasympathetic ~ mediation as the subject lies down from the upright `,'~r:~ 20 position has been well established by cumbersome and indirect techniques that assess autonomic activity.
~-- That our efficiant noninvasive method for determination ~~ of ANS responsiveness is capable of demonstrating the , ~
same result represents a surprising and important new -~ 25 adYance in the art of quantifying autonomic function.
The changes in autonomic responsiveness demonstrated in this example are, in fact, quite subtle compared ~o the dramatic alterations that frequently accompany disease.
, We therefore expect this approach to be of great utility - 30 in patient monitoring ~or diagnostic purposes and for '~ guiding therapeutic management in both critically ill and ambulatory patients.
Flg. 4 illustrates a hardware implementation of ~4 the abova-described methods. A cue generator 50 .: .
.
;-, ,;
' ```'?

., 9~tiS~

generates breathing cues and displays them to a human . subject 52. The CUQS from the cue generator 50 may be audible tones or other cues such as visible cues. The cues generated by the cua generator 50 have the 5 broad-band characteristics discussed abovQ. The human subject 52 is instrumented to provide ECG and lung volume signals on lines 54 and 56 aæ input to a computer 5~. A suitable computer is a Motorola . 68010-based computQr system. The computsr 58 computes 10 the transfer funation as described above. It should be noted that the implementation of Fig. 4 is entirely - exQmplary, and variables other than the ECG and lung voluma may be monitored as inputs to the computer 58 for transfer function analysis. Some other suitable 15 variable6 will now be described.
~; The mathods outlined in the example abova demonstrate how one might characterize autonomic behavior using one particular input signal and one form of transfer analysis. As discussed earlier, there `! 20 exists a number of hemodynamic variables sensed by the ANS that form suitable choices for input signals. For example, the ANS senses fluctuations in blood pressure . .
in th~ systemic arterial tree, the pulmonary vasculature, and the cardiac atria. Arterial pressure, 25 particularly in the systemic periphery, is easily measured either minimally invasively via a fluid-filled catheter or noninvasivaly using a plethysmographic method incorporating a finger cuff ~18]. One may thus use this pressurQ signal as the input, provided he has 30 developed a method for imposing a broad-band signal on it. This may be accomplished by modulating the air . pressure in a body suit or neck chamber worn by the subject. Other suitable input signals might include visual stimuli, such as light flashes, auditory cues, ' '' ~":.;, '`

. -.~

~L29~3651~

~;;;, and the like.
Tha broad-band waveform employed for imposition on the input signal also need not bo the same as in the ; example section above. Marmarelis and Marmarelis ~15]
- 5 have discussed the statistical properties of several classas of such waveforms, including Gaussian white noisa, random switching signal~, and pseudorandum binary signals. Each of thesQ types of broad-band waveforms ;~ have different applications, and frequently one is more easily imposed on the input variable used than the ~- others. When art0rial pressure is the input ~ariable and is modulated by a neck chamber, for instance, the random switching signal is an ideal choice. In this case, ~he neck chamber is connected to a Y-valve that toggles betw~en two constant pressure sources at random time intervals.
In some applications it may be desirable to modify the input waveform so as to enrich its spectral . content in a specific frequency band. If, ~or example, ~` 20 the cohQrence function demonstratss the presence of additivQ noise corrupting the transfer function estimates within a particular band, then enriching the input signal may improve the signal to noise ratio and thus the transfQr functions estimates in that band.
2s This sQlectivQ enrichment may be accomplished in a number of ways, such as through the USQ of ~ilters that accentuate the spectral content in some frequency bands -- and attenuate it in others. In the example section above, W8 discussed another method in which the input signal spectrum could be altered by modifying the distribution of intervals between random events.
.~` Flnally, we must emphasize that the signal processing technique utilized in the example section .,.
~ ~ above, namely computation of the transfer function `:
~....
~: &

1298~à56 magnitude and phase as a function of frequency, is by no means the only such technique available. It is a particularly useful method when no a priori knowledge about the system under study is available. When a specific model or class of models is Xnown to well describe the behavior of the system, then another signal processing algorithm may provide a more efficient means of characterizing the system than transfer function analysis. Regression techniques, for instance, provide a useful time domain approach for quanitifying system - behavior when one can assume a maximum time lag over which the value of the input signal at one point in time continues to affect the output.
Each of the fundamental components of our method for analysis of the ANS thus consists of a range . of available options. These include a wide variety of `. choices for the input variable, various broad-band waveforms for use as the input excitation, and a diverse ,~ armamentarium of signal processing techniques. One may thus tailor the generalized approach as needed depending on what aspect of autonomic behavior is being probed.

,. . .
. ~ .
. ".~ .

; .,~
.

~,...
.~, ,~
,~"~

: ,-~ ,i .:.

SJ
_ _ !365E;

- APPENDIX A
~-. 1. S. Akselrod, D. Gordon, F.A. Ubel, D.C.
Shannon, A.C. Barger, and R.J. Cohen, "Power - Spectrum Analysis of Heart Rate Fluctuations: A
~ Quantitative Probe of Beat-to-Beat ; Cardiovascular Control", Science, Vol. 213, pp.
220-222 ~1981).

2. S. Akselrod, D. Gordon, J.B. Madwed, N.C.
Snidman, D.C. Shannon, and R.J. Cohen, "Hemodynamic Regulation: Investigation by Spectral Analysis", Am. J._Physiol., (in press).
, , 3. . ~.F. Chess and F.R. Calaresu, "Frequency ;;~ Response Model of Vagal Control of Heart Rate ;~ in the Cat" Am. J. PhYsiol~~ Vol. 220, pp.
554-557 (1971).
:'~

4. ~.F. Chess, R.M.K. Tam, and F.R. Calaresu, -~. "Influence of Cardiae Neural Inputs on Rhythmic ~ Variations of Heart Period in the Cat", i ~m. J. PhYsiol., Vol. 228, pp. 775-780 (1975).

,, t' ~ 5. R.W. DeBoer, J.M. Karemaker, and J. Strackee, ~:~ "Comparing Spectra of a Series of Point Events Particularly for Heart Rate Variability Data", ' IEEE Trans. Biomed. Enq., Vol. BME-31, pp.
. ~
~; 3~4-387 (1984).
..
: i:
6. R.W. DeBoer, J.M. Karemaker, and J. StrackeQ, eat-to-~eat Variability of Heart Interval and ~ Blood Pressure", Automedica, Vol. 4, pp.
:~ 217-222 (1983).
, A~

.
~.~

.:
, ~:
' "

''~

.

1~
~9~65~i -18~ -. . .
7. R.W. DeBoer, J.M. Karemaker, and J. Strackee, "Relationship ~etween Short-term Blood-pressure Fluctuations and Heart-rate variability in Resting Subjects I: A Spectral Analysis Approach", Med. +
Biol. Enq. + ComP., Vol. 23, (1985).

8. G. N. Franz, A.M. Scher, and C.S. Ito, "Small -. Signal Characteristics of Carotid Sinus x~ , Baroreceptors o~ Rabbits", J. Applied Physiol., Vol. 30, pp. 527 535 (1971).
; .-9. B.W. Hyndman and J.R. Gregory, "Spectral Analysis of Sinus Arrhythmia During Mental Loading", Erqonomics, Vol. 18, pp. 255-270 (1975).
; ,, ~.. ' 10. R.I. ~itnay, T. Fulton, A.H. McDonald, and D.A.
.~r' Linkens, "Transient Interactions Between Blood Pressure, Respiration and Heart Rate in Man", in press, ().

:: 11. C.F. ~napp, J.M. Evans, D.C. Randall, and J.A.
Marquis, "Cardiovascular Regulation in Canines During Low-frequency Acceleration", Am. J.
PhYsiol.~ Vol. 243 pp. H998-H1009 (1982).
"~ ,;
~ , f~" 12. W.H. Levison, G.O. Barnett, and W.D. Jackson, - "Nonlinear Analysis o~ the Baroreceptor Reflex ~ System", Circ. Res., Vol. 18, pp. 673-682 (1966).
:' . 13. T.C.Lloyd, "Cardiopulmonary Baroreflexes:
- Integrated Responses to Sine- and Square-wave Forcing", J. Applied Phvsiol., Vol. 35, pp. 870-874 (1973).
.-, .;, e .. . . ... ... .

129~56 .- 14. H. Luczak and W. Laurig, "An Analysis of Heart Rate ;~. Variability", Erqonomics, Vol. 16, pp. 85-97 (1973).
~`
15. P.Z. Marmarelis and V.Z. Marmarelis, '`Analysis of Physiological Systems", Plenum Press, New Yor~
(1~7~) -~ 16. J. Penaz, N. Honzikova, and B. Fizer, "Spectral Analysis of Resting Variability of Some Circulatory Parameters in Man", Physiol. Bohem., Vol. 27, pp.
9-357 (1978).
. ~:
17. J. Penaz, "Frequency Response of the Cardiac Chronotropic Action of the Vagus in the Rabbit", Arch. Int. Physio. Bioch., Vol. 70, pp. 636-650 (1962).

~ -.`-!' 18. J. Penaz, "Photoelectric Measurement of Blood Pressure, Volume and Flow in the Finger", Diqest . 10th Int'l. Conf. Med. Bio. Enq., p. 104 (1973).
: i .
-.. : 19. B. Pomeranz, R.J.B. McCaulay, M.A. Caudill, I.
,,; ,, -Kutz, D. Adam, D. Gordon, K.M. Kilborn, A.C.
~arger, D.C. Shannon, R.J. Cohen, and H. Benson, "Assessment of Autonomic Function in Humans by Heart Rate Spectral Analysis", Am. J. PhYsiol., Vol. 248, pp. H151-H153 (1985~.
' ~ .
20. A.M. Scher, W. W. Ohm, K. Bumgarner, R. Boynton, - and A.C. Young, "Sympathetic and Parasympathetic Control of Heart Rate in the Dog, Baboon and Man", ~ .
il~ Fed. Proc., Vol 31, pp. 1219-1225 (1972).
, ~;.. ,~ .
~i . 2.;~

, ;: _ ~ A ~
.
':
''~$

, ~ - - '"- '''`
,.

~2~65~i ~20-.~x 21. A.M. Scher and A. C . Young, "Servoanalysis of '','A` Carotid Sinus ~eflex Effects on Peripheral .. ..
~- Resistance", Cir. Res., Vol. 12, pp. lS2-162 (1963).
., 22. A.M. Scher and A.C. Young, "Reflex Control of Heart Rate in the Unanesthetized Dog", Am. J. Physiol., Vol. ~18, pp. 780-789 (1970).

- 23. J.W. Spickler, P. Kezdi, and E. Geller, "Transfer Characteristics o the Carotid Sinus Pressure - Control System", pp. 31-40 in Baroreceptors and Hypertension, ed. P. Kezdi, Pergamon Press, Oxford (19~7).

24. M.G, Taylor, "Use of Random Excitation and Spectral Analysis i~ the Study of Frequency-dependent Parameters of the Cardiovascular System", Circ.
Res., Vol. 18, pp. 585-595 (1966).
" ~
25. U. Zwiener, "Physiological Interpretation of -- Autospectra, Coherence and Phase Spectra of Blood Pressure, Heart Rate, and Respiration Waves in Man", Automedica, Vol. 2, pp. 161-169 t1978)-:

. ,~, .
, .
..., j;
;-;, ., -.

v~
_ _ _ _ _ _ .,,, ,.,~ . ,~.

Claims (33)

1. Method for assessing autonomic response comprising:
applying a broad-band perturbation to a physiologic input signal sensed by the autonomic nervous system;
monitoring a physiologic output signal modulated by the autonomic nervous system; and computing the transfer relation between the physiologic input and output signals.

2. The method of claim 1 wherein the input signal to which the broad-band perturbation is applied is respiratory activity.

3. The method of claim 1 wherein the physiologic output signal is heart rate.

4. The method of claim 1 in which the broad-band perturbation has a Poisson distribution.

5. The method of claim 1 wherein the broad-band perturbation has a Poisson distribution modified to prohibit intervals shorter than some minimum interval tmin.

6. The method of claim 1 wherein the broad-band perturbation is a Poisson distribution modified to prohibit intervals shorter than some minimum interval tmin or longer than some maximum interval tmax.

7. The method of claim 1 wherein the broad-band perturbation lasts for approximately 6 minutes.

8. The method of claim 1 wherein the broad-band perturbation contains components in the range of 0.0 - 0.5 Hz.

9. The method of claim 1 wherein the transfer relation is a complex transfer function.

10. The method of claim 9 wherein the complex transfer function is decomposed into magnitude and phase components.

11. Method for assessing autonomic response comprising:
applying a broad-band perturbation to respiratory activity which is the input signal sensed by the autonomic nervous system;
monitoring heart rate as the output signal modulated by the autonomic nervous system; and computing the transfer relation between respiratory activity and heart rate.

12. The method of claim 1 wherein the autonomic response is assessed on a standing subject.

13. The method of claim 1 wherein the autonomic response is assessed on a supine subject.

14. The method of claim 1 wherein the broad-band perturbation is applied to arterial pressure as the input signal.

15. The method of claim 14 wherein arterial pressure variations are accomplished by modulating air pressure in a body suit or neck chamber worn by a subject.

16. The method of claim 1 wherein the broad-band perturbation has a Gaussian white noise distribution.

17. The method of claim 1 wherein the broad-band perturbation has a random switching signal distribution.

18. The method of claim 1 wherein the broad-band perturbation has a pseudo random binary distribution.

19. The method of claim 1 wherein the broad-band perturbation is modified in a specific frequency band.

20. Method for assessing autonomic response comprising;
applying a broad-band perturbation to a physiologic parameter which is the input signal sensed by the autonomic nervous system;
monitoring an output signal which is a physiologic parameter which is modulated by the autonomic nervous system; and computing the relationship between the input and output signals.

21, The method of claim 20 wherein the relationship between the input signal and the output signal is performed by a regression analysis.

22. Apparatus for assessing autonomic response comprising means for subjecting a physiologic parameter sensed by the autonomic system as an input signal to a broad-band perturbation;
means for monitoring an output signal which is a physiologic parameter modulated by the autonomic nervous system; and means for computing the relationship between the input and output signals.

23. The apparatus of claim 22 wherein the broad-band perturbation is applied to respiratory activity as the input signal.

24. The apparatus of claim 22 wherein the physiologic output signal is heart rate.

25. The apparatus of claim 22 wherein the relationship between the input signal and the output signal is a transfer function.

26. The apparatus of claim 22 wherein the broad-band perturbation applied to respiratory activity as the input signal is induced by external cues to which the subject is trained to respond.

27. The apparatus of claim 22 wherein the broad-band perturbation has a Poisson distribution.

28. The apparatus of claim 22 wherein the relationship between the input signal and the output signal is determined by a regression analysis.

29. The method of claim 1 wherein the physiologic output signal is a hemodynamic parameter.

30. The apparatus of claim 22 wherein the output signal is a hemodynamic parameter.

31. The method of claim 1 wherein the perturbation is auditory stimuli.

32. The method of claim 1 wherein the perturbation is visual stimuli.

33. The method of claim 32 wherein the visual stimuli are light flashes.