1: \documentclass[12pt]{article}
2: \usepackage{epsfig}
3: %\usepackage{graphicx}
4: %\topmargin=+0.0in
5: %\oddsidemargin=0.25in
6: \textwidth=6.0in
7: %\textheight=8in
8: %\footskip=6ex
9: %\footheight=2ex
10:
11: \begin{document}
12:
13: \title{Nonstationary time series analysis of\\
14: heart rate variability}
15: \author{P.B. Siegel$^1$, J. Sperber$^2$, W. Kindermann$^2$,
16: and A. Urhausen$^2$}
17: \maketitle
18: \noindent $^1$Physics Department, California State
19: Polytechnic University Pomona, Pomona, CA 91768\\
20: $^2$Institute of Sport and Preventative Medicine, University of
21: Saarland, 66041 Saarbruecken, Germany
22:
23:
24: \begin{abstract}
25: An analysis of the RR-interval time series, $t_i$, is presented for
26: the case in which the average time, $\bar{t}$, changes slowly.
27: In particular, $\bar{t}$ and a short-time scale variability
28: parameter, $V$, are simultaneously measured while $\bar{t}$ decreases
29: for subjects in the reclined position.
30: The initial decrease in $\bar{t}$ is usually linear with
31: $V$ yielding parameters that can be related to
32: physiological quantities.
33: \end{abstract}
34:
35: \centerline{PACS: 87.19.Hh, 87.80.Tq, 87.80.Vt, 05.45.Tp}
36:
37: \section{Introduction}
38:
39: Heart rate variability is used to examine the
40: autonomous nervous system's control of heart rate, with the goal
41: of assessing health and fitness in humans\cite{eckberg}. The main
42: measurement is the time between successive heartbeats, which
43: is referred to as the RR-interval. Measurements are taken
44: over time periods that can be as short as minutes or as
45: long as days, and the resulting data are a series of times, $t_i$,
46: measured to an accuracy of milliseconds. Data analysis includes
47: methods from spectral analysis\cite{prok, eckberg2, malik, pagani},
48: statistical physics
49: \cite{costa, yulm}, and nonlinear dynamics\cite{peng, thurner, wessel, ching}.
50: These methods are applied to situations for which the time
51: series is stationary, that is,
52: the average value of the $t_i$ does not change significantly in time.
53: These approaches have been useful
54: in determining certain aspects of heart rate control. However, there may
55: be ambiguities in relating the analysis parameters to physiological quantities.
56:
57: The purpose of this article is to consider the feasibility of using a
58: controlled nonstationary time series
59: of the RR-interval times to assess heart rate control. If
60: the average value of the $t_i$, $\bar{t}$, changes sufficiently
61: slowly, then it
62: is possible to also measure a short-time scale variability
63: parameter, $V$.
64: If $\bar{t}$ and $V$ depend upon physiological
65: quantities in different ways, then a plot in the $\bar{t} - V$
66: plane for a controlled process may yield information about the
67: influences causing the change. Although we use a specific
68: measurement for $V$, the respiratory sinus arrhythmia amplitude,
69: the approach can be applied for any measure of $V$.
70:
71:
72: \section{Description of Measurements and Data Analysis}
73:
74: The average heart period ($\bar{t}$) and the variability ($V$) of the
75: RR-interval times depend primarily on the
76: autonomous nervous system, which has two main
77: influences: sympathetic ($m$) and parasympathetic ($n$) nerve activity.
78: One would like to learn about $m$ and $n$ from the non-invasive
79: measurements of $\bar{t}$ and $V$. A model used
80: by physiologists to describe the relationship of the average heart
81: rate $B$ to the sympathetic and parasympathetic influences is given by
82: the equation \cite{katona, fouad, kenney} $B = B_0 m n$
83: where $B_0$ is the intrinsic heart rate, $m$ is the sympathetic
84: factor and $n$ the parasympathetic factor. The sympathetic
85: factor $m$ is greater than one, increasing the heart rate, and
86: the parasympathetic factor $n$ is less than one, decreasing the
87: heart rate. The factors $m$ and $n$ which multiply the intrinsic heart rate
88: depend, among other influences, upon a person's activity level, posture, recent
89: diet, degree of fitness, age, and health. We label all
90: these parameters, which affect the sympathetic (m) and parasympathetic (n)
91: factors, by the symbol $\beta$: m($\beta$) and n($\beta$). One can
92: also write $n=1/(1+P(\beta))$ \cite{katona}, where the parasympathetic
93: control parameter $P$ is unitless and varies from zero to its maximum
94: value $P_{max}$. In Ref. \cite{katona} $P$ has units of milli-seconds.
95: In terms of the average heart period, the equation can
96: be written as $\bar{t} = T_0(1+P(\beta))/m(\beta)$, where
97: $T_0 = 1/B_0$ is the intrinsic heart period.
98:
99: While in general variability is influenced by both sympathetic and
100: parasympathetic effects, short-time scale variability depends primarily
101: on parasympathetic activity. Short-time scales are
102: variations with frequencies greater than $0.15$ Hz. Since
103: average heart rates are around one beat per second, this corresponds
104: to variations in heart rate taking place in less than roughly 6 heartbeats.
105: There is good evidence that high frequency ($>0.15$Hz) spectral power
106: depends only on $P$, and is not influenced by sympathetic nerve
107: activity. Thus, if $V$ is a measure of short-time
108: scale variation, it can be expressed as a function of parameters
109: other than $m$, namely $V(P, \beta)$.
110: Expressing $P$ in terms of $V$ and the "$\beta$" factors,
111: the relationship between the
112: average heart period and its variability can be written as
113: $\bar{t} = T_0 (1+P(V, \beta))/m(\beta)$. Measurements
114: of $\bar{t}$ and $V$ during slow changes of these values can
115: yield information about changes in $m$, $T_0$, and the
116: function $P(V, \beta)$.
117:
118: For this study, we chose as the variability parameter $V$ the
119: respiratory sinus arrhythmia (RSA) amplitude,
120: which is the amplitude of the fluctuations of the RR-interval times
121: due to breathing, for a breathing rate of 12 breaths/min.
122: We have chosen this parameter for $V$ since it has been shown in
123: unconscious dogs\cite{katona} and humans\cite{fouad} that
124: ($\bar{t} - T_0/m$) is proportional to the RSA amplitude.
125: The experiments carried out in Refs. \cite{katona} and \cite{fouad}
126: used pharmacological methods (parasympathetic blockade) to decrease $\bar{t}$
127: while keeping $m$ constant. If this linear relationship
128: holds in general, then $V(P, \beta)$ takes on the simple
129: form $V = k(\beta)P$, where k has the same units as $V$.
130: Letting $V$ represent the RSA amplitude for a 12 breath/min breathing
131: rate, we have
132:
133: \begin{equation}
134: \bar{t} = {T_0 \over m} (1 + {{V} \over {k}} )
135: \end{equation}
136:
137: \noindent where $m$ and $k$ might depend on posture, tidal volume,
138: environmental factors, diet before measurement, etc. The hope is that
139: $\bar{t}$ and $V$ can be varied keeping $k$ constant.
140:
141: We measured both the average heart period in msec, and the amplitude
142: of the RSA in
143: msec for various subjects while they underwent slow warm up in the lying
144: position. Since the amounts of sympathetic and parasympathetic influences
145: are posture dependent, it is important to maintain the same posture
146: while the heart rate changes. The lying position was chosen because
147: the parasympathetic activity is maximized in this position.
148: The subject slowly increased his/her pulse
149: by pedaling a stationary bicycle in the lying position while breathing
150: at 12 breaths/min the whole time. He/she started by lying still
151: for 2 minutes, and then began pedaling slowly with a load of 25
152: watts for the duration of one minute. The subject then pedaled faster until
153: the heart rate increased to around 20 bpm above its resting value. The
154: load was increased by increments of 25 watts each two minutes
155: thereafter, until a pulse of around 110 beats/min was reached.
156:
157: Breathing was controlled by having the subjects breathe in concert
158: with increasing and decreasing tones, which they listened to from a
159: CD player. RR interval times were recorded using a heart rate
160: monitor from Polar, Model S810. The average heart period
161: and RSA amplitude were computed from a
162: sequence of 100 consecutive RR-interval times, with the start of the
163: interval shifted 50 beats after each calculation. For example, the first
164: sequence is beat numbers 1 to 100, the second from 50 to 150, etc.
165: Since breathing was
166: controlled at 12 breaths/min, the peak in the Fourier spectrum
167: was distinct at 0.2 Hz.
168: From each sequence of 100 times, the RSA amplitude was set equal to this
169: peak Fourier amplitude at or near the breathing frequency of
170: 12 cycles/min (i.e. 0.2 Hz).
171:
172: \section{Discussion}
173:
174: We carried out the test for 20 subjects, and plot some
175: typical cases in figures 1 to 3. In each figure, the initial part of the warm
176: up is linear, labelled $1 \rightarrow 2$, in
177: which both the heart period and the variability decrease.
178: This is followed by a decrease in heart period with little change
179: in variability, labelled as $2 \rightarrow 3$ in the figures. About
180: $75\%$ of the subjects we measured had similar plots.
181:
182: The data can be interpreted using the model of Eq. 1. During the initial linear
183: stage of the warm-up, $1 \rightarrow 2$, parasympathetic influences
184: are reduced since the RSA amplitude decreases. If the RSA amplitude
185: is proportional to $P$, the data are consistent with
186: $T_0/m$ and $k$ being constant during this first stage. After the
187: initial parasympathetic decrease,
188: the plot moves to the left, $2 \rightarrow 3$. Since the variability does
189: not change during this stage, the data indicate that $m$ is increasing.
190: For the initial linear stage, three parameters can
191: be determined: $T_0/m$, $k$, and $P_{max}$. The parameter $P_{max}$
192: is unitless,
193: and is the fraction that the parasympathetic activity increases the
194: heart period from $T_0$ to the maximum value of $\bar{t}$:
195: $P_{max} = (t_{max} - T_0/m)/(T_0/m)$, where
196: $t_{max}$ is the maximun value of the average heart period.
197: These parameters can always be determined for a linear stage of the graph.
198: If the model of Eq. 1 is valid and $k$ remains constant, then
199: the parameters have physiological significance.
200: In the figures we state $P_{max}$, $k$, and $T_0/m$ for the
201: slow warm-up phase.
202:
203: Some subjects show interesting results, which we graph in figures 2 and 3. In
204: figure 2, the plot shows four stages which can be interpreted using Eq. 1.
205: Initially ($1 \rightarrow 2$) the RSA amplitude decreases significantly
206: suggesting $P$ decreases with $m$ constant. Then ($2 \rightarrow 2a$)
207: the heart period decreases but the RSA amplitude remains constant, meaning
208: that $m$ increases with $P$ constant. The next change from $2a \rightarrow 2b$
209: is similar to the change from $1 \rightarrow 2$, and the final stage from
210: $2b \rightarrow 3$ has $m$ increasing with $P$ constant as from
211: $2 \rightarrow 2a$. For this subject it appears that the heart rate
212: increases via alternating parasympathetic and sympathetic influences.
213:
214: In figure 3 we compare two different subjects. Both subjects have the
215: same resting heart rate of 50 bpm. One subject (circles) has a resting RSA amplitude
216: of 70 msec, while the other (triangles) only 25 msec. One might be lead to believe
217: that the first subject has a stronger parasympathetic activity, since the
218: RSA amplitude is larger. However, during warm-up the RSA amplitude for
219: the first subject (circles) was reduced to its minimum value of 10 msec
220: for only a 350 msec decrease in
221: heart period. For the other subject (triangles), the heart period
222: decreased by 600 msec
223: before the RSA amplitude was reduced to its minimum value. Thus the
224: second subject has a larger change in heart period due to the reduction
225: in parasympathetic activity, and consequently a larger value of $P_{max}$.
226: The plots in the $\bar{t}-V$ plane are quite different for these
227: subjects even though they have the same resting heart rate.
228:
229: We also used the absolute value of the difference between successive
230: RR-intervals without controled breathing as a variability parameter.
231: If the subject is breathing
232: at a consistent rate throughout the exercise, the results are
233: similar to those obtained using the RSA amplitude
234: with a different value of $k$. Other possible
235: candidates for a variability parameter are the standard deviation of the
236: RR-interval times, the square-root of the
237: high-frequency power, or the RSA from Ref. \cite{hatfield}.
238: Heart rate ($1/\bar{t}$) versus
239: variability are plotted in Refs. \cite{hatfield} and \cite{grossman}.
240: However, in neither case are the data modeled using the parameters of
241: Eq. 1. In figure 4 of Ref. \cite{grossman}, the loci of points are similar
242: to the figures presented here: a large decrease in RSA for a small increase
243: in heart rate (parasympathetic reduction), and a small decrease in
244: RSA for a large increase in heart rate (sympathetic increase). Perhaps
245: if the time ordering of the points and posture are taken into account,
246: $T_0/m$, $P_{max}$, and $k$ could be measured for different types of daily
247: activity.
248:
249: In conclusion, we propose that a simultaneous measurement of the
250: average heart period $\bar{t}$ and a parameter related to short time-scale variability
251: $V$ can yield insight into the autonomous nervous system. For slow
252: changes in the heart rate, a graph in the $\bar{t}-V$ plane can assist
253: in understanding the influences causing the change. For situations
254: in which the change in heart rate is caused only by the parasympathetic
255: nerve activity, one can extract three parameters, $T_0/m$, $P_{max}$,
256: and $k$. These parameters could have a simple relationship to
257: physiological quantities.
258:
259: \begin{thebibliography}{hghg}
260:
261: \bibitem{eckberg} D. Eckberg, Ann. Med. {\bf 32}, 341-349 (2000).
262:
263: \bibitem{prok} M.D. Prokhrov, V.I. Ponomarenko, V.I. Gridnev, M.B. Bodrov,
264: and A.B. Bespyatov, Phys. Rev. E {\bf 68}, 041913 (2003).
265:
266: \bibitem{eckberg2} D. Eckberg, Circulation {\bf 96}, 3224-3232 (1997).
267:
268: \bibitem{malik} M. Malik, Circulation {\bf 93}, 1043-1065 (1996).
269:
270: \bibitem{pagani} M. Pagani et. al., Circulation Research {\bf 59},
271: 178-193 (1986).
272:
273: \bibitem{costa} M. Costa, A.L. Goldberger, and C.-K. Peng, Phys.
274: Rev. Lett. {\bf 89} 068102 (2002).
275:
276: \bibitem{yulm} R. Yulmetyev, P. Hanggi, and F. Gafarov, Phys. Rev.
277: E {\bf 65}, 022901 (2002).
278:
279: \bibitem{peng} C.-K. Peng, J. Mietus, J>M> Hausdorff, S. Havlin, H.E.
280: Stanley, and A.L. Goldberger, Phys. Rev. Lett. {\bf 70}, 1343-1346 (1993).
281:
282: \bibitem{thurner} S. Thurner, M.C. Feurstein, and M.C. Teich, Phys.
283: Rev. Lett. {\bf 80}, 1544-1547 (1998).
284:
285: \bibitem{wessel} M. Wessel, A. Schirdewan, and J. Kurths, Phys. Rev.
286: Lett. {\bf 91}, 119801 (2003).
287:
288: \bibitem{ching} E. Ching, D.C. Lin, and C. Zhang, Phys. Rev. E
289: {\bf 69}, 051919 (2004).
290:
291: \bibitem{katona} P. Katona and F. Jih, J. Appl. Physiol. {\bf 39},
292: 801-805 (1975).
293:
294: \bibitem{fouad} F. Fouad, R. Tarazi, C. Ferrario, C Fighaly
295: and C. Alicandri, Am. J. Physiol. {\bf 246}: H838-H842 (1984).
296:
297: \bibitem{kenney} W. Kenney, Med. Sci. Sports. Exer. {\bf 17},
298: 451-455 (1985).
299:
300: \bibitem{hatfield} B. Hatfield, T. Spalding,
301: D. Santa Maria, S. Porges, J. Potts, E. Byrne, E. Brody, and
302: A. Mahon, Med. Sci. Sports Exerc. {\bf 206}, 14 (1998).
303:
304: \bibitem{grossman} P. Grossman, F. Wilhelm, and M. Spoerle, Am. J.
305: Physiol.Heart Circ. Physiol. PMID: 14751862 (2004).
306:
307: \end{thebibliography}
308:
309: \newpage
310:
311: \begin{center}
312: {\bf Figure Captions}
313: \end{center}
314:
315: \bigskip
316:
317: \noindent Figure 1. Graph of RSA (respiratory sinus arrhythmia)
318: versus heart period as a subject's heart period slowly deceases while
319: in the lying position. From the straight-line section $1 \rightarrow 2$
320: the parameters $T_0/m$, $P_{max}$, and $k$ are determined using equation 1.
321:
322: \bigskip
323:
324: \noindent Figure 2. Same graph as in Figure 1, but for a different
325: subject. The decrease in heart period can be interpreted as alternating
326: between the influences of parasympathetic decrease ($1 \rightarrow 2$)
327: and ($2a \rightarrow 2b$) and sympathetic increase ($2 \rightarrow 2a$)
328: and ($2b \rightarrow 3$).
329:
330: \bigskip
331:
332: \noindent Figure 3. Same graph as in Figure 1, comparing two different
333: subjects. Both subjects have the same resting rate. Eventhough one
334: subject (circles) has a higher RSA, the parasympathetic control
335: parameter $P_{max}$ is less.
336:
337: \newpage
338:
339: \begin{figure}
340: \epsfxsize=16cm
341: \epsfbox{fig1siegel.ps}
342: \end{figure}
343:
344: \noindent Figure 1. Graph of RSA (respiratory sinus arrhythmia)
345: versus heart period as a subject's heart period slowly decreases while
346: in the lying position. From the straight-line section $1 \rightarrow 2$
347: the parameters $T_0/m$, $P_{max}$, and $k$ are determined using equation 1.
348:
349: \newpage
350:
351: \begin{figure}
352: \epsfxsize=16cm
353: \epsfbox{fig2siegel.ps}
354: \end{figure}
355:
356: \noindent Figure 2. Same graph as in Figure 1, but for a different
357: subject. The decrease in heart period can be interpreted as alternating
358: between the influences of parasympathetic decrease ($1 \rightarrow 2$)
359: and ($2a \rightarrow 2b$) and sympathetic increase ($2 \rightarrow 2a$)
360: and ($2b \rightarrow 3$).
361:
362: \newpage
363:
364: \begin{figure}
365: \epsfxsize=16cm
366: \epsfbox{fig3siegel.ps}
367: \end{figure}
368:
369: \noindent Figure 3. Same graph as in Figure 1, comparing two different
370: subjects. Both subjects have the same resting rate. Eventhough one
371: subject (circles) has a higher RSA, the parasympathetic control
372: parameter $P_{max}$ is less.
373:
374: \end{document}
375:
376: