q-bio0409039/kiyono.tex
1: \documentclass[aps,amsmath,amssymb,prl,twocolumn,showpacs,superscriptaddress]{revtex4}
2: \usepackage{graphicx}
3: 
4: \begin{document}
5: 
6: \title{Critical Scale-invariance in Healthy Human Heart Rate}
7: 
8: \author{Ken Kiyono}
9: \affiliation{Educational Physiology Laboratory, Graduate School of
10:   Education, The University of Tokyo, 7--3--1 Hongo, Bunkyo-ku, Tokyo
11:   113--0033, Japan}
12: \author{Zbigniew R. Struzik}
13: \affiliation{Educational Physiology Laboratory, Graduate School of
14:   Education, The University of Tokyo, 7--3--1 Hongo, Bunkyo-ku, Tokyo
15:   113--0033, Japan}
16: \affiliation{PRESTO, Japan Science and Technology Agency, Kawaguchi,
17:   Saitama 332--0012, Japan}
18: \author{Naoko Aoyagi}
19: \affiliation{Educational Physiology Laboratory, Graduate School of
20:   Education, The University of Tokyo, 7--3--1 Hongo, Bunkyo-ku, Tokyo
21:   113--0033, Japan}
22: \author{Seiichiro Sakata}
23: \author{Junichiro Hayano}
24: \affiliation{Core Laboratory, Nagoya City University Graduate School
25:   of Medical Sciences, 1 Kawasumi, Mizuho-cho, Mizuho-ku, Nagoya
26:   467--8601, Japan}
27: \author{Yoshiharu Yamamoto}
28: \email{yamamoto@p.u-tokyo.ac.jp}
29: \affiliation{Educational Physiology Laboratory, Graduate School of
30:   Education, The University of Tokyo, 7--3--1 Hongo, Bunkyo-ku, Tokyo
31:   113--0033, Japan}
32: \affiliation{PRESTO, Japan Science and Technology Agency, Kawaguchi,
33:   Saitama 332--0012, Japan}
34: 
35: \date{\today}
36: 
37: \begin{abstract}
38: We demonstrate the robust scale-invariance in the probability
39: density function (PDF) of detrended healthy human heart rate
40: increments, which is preserved not only in a quiescent condition,
41: but also in a dynamic state where the mean level of heart rate is
42: dramatically changing. This scale-independent and fractal
43: structure is markedly different from the scale-dependent PDF
44: evolution observed in a turbulent-like, cascade heart rate model.
45: These results strongly support the view that healthy human heart
46: rate is controlled to converge continually to a critical state.
47: \end{abstract}
48: \pacs{87.19.Hh, 05.40.-a, 87.80.Vt, 89.75.Da}
49: 
50: \maketitle
51: 
52: Healthy human heart rate belongs to a special class of complex
53: signals, showing long-range temporal correlations
54: \cite{1357,1430}, non-Gaussianity of the increment's PDF
55: \cite{1357} and multifractal scaling properties \cite{1620,Amaral01}, 
56: and has served as the `benchmark' of choice for studies of biological
57: complexity. Two alternative mechanisms, both characterized by these
58: three features, have been proposed for this heart rate complexity:
59: 1) a random (multiplicative) cascade process based on the resemblance
60: of the behavior of the structure function \cite{Frisch} of heart rate
61: increments to that of spatial velocity differences in hydrodynamic
62: turbulence \cite{1759}; and 2) critical state-like dynamics \cite{1522}
63: based on the resemblance of the scale-invariant properties in heart
64: rate to those of many systems operating near the critical point of
65: their phase space. To date the exact mechanism for the complex heart
66: rate dynamics is unknown. However, as it reflects the dynamics of the
67: autonomic nervous system's control of heart rate \cite{1357,Amaral01}
68: and thus provides potential predictors for the mortality of cardiac
69: patients \cite{1532,1705,1305}, elucidating this mechanism is
70: considered important.
71: 
72: Lin and Hughson \cite{1759} recently reported an analogy between
73: turbulence and human heart rate dynamics by finding similarity of the
74: structure function---directly linked with multifractal formalism
75: \cite{Muzy93a}---of heart rate increments to that of spatial velocity
76: differences in a random cascade process proposed as a model of
77: hydrodynamic turbulence. One of the common features of such
78: cascade-type multifractal models is the evolution in the shape of the
79: PDF of the increments from Gaussian at large scales to stretched
80: exponential at smaller scales \cite{Castaing90}. In this study, we
81: disprove this cascade-like assertion and demonstrate that heart rate
82: signals do not follow the evolution in the shape of the (increment)
83: PDF characteristic for cascade-like processes, but report for the
84: first time a {\em robust\/} scale invariance. As long-range correlation,
85: non-Gaussianity and multifractality are also typical characteristics
86: of a system at the critical point \cite{Hinrichsen02,Kadanoff89},
87: and fluctuations in a system at a critical point are generally
88: associated with the scale invariance and universal behavior of the
89: scaling function \cite{Stanley99,Sethna2001}, we conclude that such
90: robust scale invariance in the increment PDF suggests the alternative
91: scenario of the near critical state-like operation for the healthy
92: heart rate dynamics.
93: 
94: The long-term heart rate data analyzed have been measured as
95: sequential heart interbeat intervals $b(i)$, where $i$ is the beat
96: number. We investigate the PDF of heart rate increments at
97: different time scales (in beat numbers), where the
98: non-stationarity of the data has been eliminated by local
99: detrending \cite{1357}. We first integrate the $b(i)$,
100: $B(m)=\sum_{j=1}^{m} b(j)$, and the resultant $B(m)$ is divided
101: into sliding segments of size $2n$. Then in each segment
102: the best $q$-th order polynomial is fit to the data. The
103: differences $\Delta_s B(i) = B^{*}(i+s) - B^{*}(i)$ at a scale
104: $s$ are obtained by sliding in time over the segments,
105: where $B^{*}(i)$ is a deviation from the polynomial fit. By this
106: procedure, the $(q-1)$-th order polynomial trends are eliminated
107: and we analyze the whole PDF of $\Delta_s B(i)$.
108: 
109: Using this method, we analyze two experimental and two
110: computer-generated data sets. The first data set consists of
111: daytime (12:00--18:00 hrs) heart rate data with a length of up to
112: $4\times 10^4$ heartbeats from 50 healthy subjects (10 females
113: and 40 males; ages $21-76$ years) without any known disease
114: affecting the autonomic control of heart rate [Fig.~\ref{fig:fig1}(a)].
115: Details of the recruitment of the subjects, screening for medical
116: problems, protocols and the data collection are described in
117: Sakata {\it et al.\/} \cite{Sakata99}. The data were collected
118: during normal daily life. The second experimental data set
119: consists of data of seven 26-hour long periods (up to $10^5$ beats),
120: collected when the subjects (7 males; ages $21-30$ years) underwent
121: `constant routine' (CR) protocol, where known behavioral factors
122: affecting heart rate (e.g. exercise, diet, postural changes and sleep)
123: are eliminated [Fig.~\ref{fig:fig1}(b)] \cite{Amaral01,Aoyagi03}.
124: 
125: \begin{figure}[tb]
126: \begin{center}
127:  \includegraphics[width = 0.92\linewidth]{fig1.eps}
128: \end{center}
129: \caption{
130: (a) A representative record of daytime (12:00--18:00 hrs) heart
131: interbeat intervals for a healthy subject. (b) An example of heartbeat 
132: intervals for a healthy subject during constant routine protocol
133: \cite{Amaral01,Aoyagi03}. (c) The surrogate data for (b).
134: (d) Data generated by a cascade heart rate increment model
135: \cite{1759}. The parameters used for simulation are $J=15$,
136: $R_t = 2$ and $\sigma_j = R_t^{-2.5-j/J}$ (see Ref.~\cite{1759}
137: for details). 
138: }
139: \label{fig:fig1}
140: \end{figure}
141: 
142: In order to test the possible presence of non-linear mechanisms
143: in complex heart rate dynamics, we apply the surrogate data test
144: to the CR protocol data  \cite{Schreiber00}. We generate a
145: surrogate data set with the same Fourier amplitudes and 
146: distributions as the original increments in the CR protocol data
147: [Fig.~\ref{fig:fig1}(c)]. Since only linear temporal correlation
148: of $b(i+1)-b(i)$ is retained in the surrogate data, a comparison
149: with the raw data can be used to test whether the PDF of
150: `velocity' increments $\Delta_s B(i)$ possesses some non-linear
151: mechanism inherent to it. Finally, we generate heart rate
152: increments of comparable lengths, following the `turbulence-like'
153: scenario from the random cascade model proposed recently by Lin
154: and Hughson [Fig.~\ref{fig:fig1}(d)] \cite{1759}.
155: 
156: PDF's of $\Delta_s B(i)$ for healthy humans, which are standardized by
157: dividing the heart rate increments in each record by the standard
158: deviation, are non-Gaussian in shape for a wide range of scales
159: $8 \leq s \leq 4,096$ irrespective of whether the subjects were
160: in their normal daily routine [Fig.~\ref{fig:fig2}(a)] or in CR
161: [Fig.~\ref{fig:fig2}(b)]. In contrast, the PDF's of the surrogate
162: data are near Gaussian, although non-Gaussianity with the fat tails
163: close to those in the observed data is still encountered at fine
164: scales [Fig.~\ref{fig:fig2}(c)]. The difference between the healthy
165: human and surrogate data indicates that the observed non-Gaussian
166: behavior is related to non-linear features of the healthy heart
167: rate dynamics. The PDF's of the cascade model show continuous
168: deformation and the appearance of fat tails when going from large
169: to small scales [Fig.~\ref{fig:fig2}(d)]. 
170: 
171: \begin{figure*}[tb]
172: \begin{center}
173:  \includegraphics[width = 0.82\linewidth]{fig2.eps}
174: \end{center}
175: \caption{
176: Deformation of increment PDF's across scales. Standardized PDF's
177: (in logarithmic scale) of $\Delta_s B(i)$ for different time
178: scales are shown for (from top to bottom) $s$ = 8, 16, 32, 64,
179: 128, 256, 512, 1024 beats. These PDF's are estimated from all
180: samples in each group. The dashed line is a Gaussian PDF for
181: comparison. (a) The PDF's from daytime (12:00--18:00 hrs) heart
182: rate time series from healthy subjects. (b) From healthy
183: subjects during constant routine protocol. (c) From surrogate
184: time series for (b). (d) From a cascade model. Note
185: in the last case (d) the continuous deformation and the
186: appearance of fat tails when going from large to fine scales.
187: In solid lines, we superimpose the deformation of the PDF using
188: Castaing's equation with the log-normal self-similarity kernel,
189: providing an excellent fit to the data.
190: }
191: \label{fig:fig2}
192: \end{figure*}
193: 
194: For a quantitative comparison, we fit the data to the following 
195: function based on Castaing's equation \cite{Castaing90}: 
196: \[
197: \tilde{P}_s(x) = \int P_{L} \left( \frac{x}{\sigma} \right)\frac{1}{\sigma}
198: G_{s,L}(\ln \sigma) {\rm d} (\ln \sigma), 
199: \]
200: where $P_{L}$ is the increment PDF at a large scale $L>s$, and
201: the self-similarity kernel $G_{s,L}$ determines the nature of
202: the cascade-type multiplicative process. Here we assume $P_{L}$
203: and $G_{s,L}$ are both Gaussian, 
204: \[
205: G_{s,L}(\ln \sigma) = \frac{1}{\sqrt{2 \pi} \lambda}
206: \exp \left(- \frac{\ln ^2 \sigma}{2 \lambda^2} \right), 
207: \]
208: and investigate the scale dependence of $\lambda^2$. The fit of
209: the PDF of actual heart rate increments to Castaing's equation
210: is indeed almost perfect, especially within $\pm 3$ times standard
211: deviation, even for a single record [Fig.~\ref{fig:fig3}(a)], and
212: robust in terms of the effect of the order of detrending
213: polynomials on the estimation of $\lambda^2$, if the order is
214: greater than two [Fig.~\ref{fig:fig3}(b)]. In the following, we use
215: the third order detrending for the estimation of $\lambda^2$.
216: 
217: Within the turbulent cascade picture, the parameter $\lambda^2$
218: can be interpreted as being proportional to the number of cascade
219: steps and is known to decrease linearly with $\log s$
220: \cite{Castaing90,Ghashghaie96}. The cascade heart rate model
221: studied here \cite{1759} also shows this effect [Fig.~\ref{fig:fig3}(c)].
222: In contrast, the scale dependency of $\lambda^2$ for healthy heart
223: rate increments is remarkably different [Fig.~\ref{fig:fig3}(c)].  
224: Especially during CR, we cannot see any decrease in $\lambda^2$
225: with $\log s$. There is no significant difference in the average
226: $\lambda^2$ at different scales, tested by the analysis of variance,
227: over the range of 23--2,048 beats for healthy subjects during daily
228: routine and 8--4,096 beats during CR [$F(13,686)=1.73$ and
229: $F(18,114)=1.70$, respectively, $p>0.05$], and the slopes of
230: $\lambda^2$ vs. $\log s$ are much closer to zero, which means the
231: absence of cascade steps across the scales in the corresponding range.  
232: 
233: In addition, when the PDF's at different scales are superimposed
234: [Fig.~\ref{fig:fig3}(d)], all the data collapse on the same curve,
235: which is one of the characteristic features observed in fluctuations
236: at a critical point \cite{Mehta02}.
237: The range of scales where this scale-invariance of PDF is observed,
238: spanning from about $10$ beats to a few thousand heartbeats, is
239: compatible with that of the robust, behavioral-independent $1/f$
240: scaling \cite{Aoyagi03} and multifractality \cite{Amaral01} of
241: heart rate. The scale-invariance in the PDF is also robust in the
242: sense that it is observed not only during CR but also during normal
243: daily life, where behavioral modifiers of heart rate 
244: dramatically change the mean level of heart rate [e.g.
245: Fig.~\ref{fig:fig1}(a)]. We thus find a novel property of scale
246: invariance in healthy human heart rate dynamics, reminiscent of
247: systems in a critical state. In particular, the invariance discovered
248: strongly supports the view that the healthy human heart is controlled
249: to converge continually to a critical state. Such a critical point
250: itself may, however, be shifted by the effects of the external and/or
251: internal environment. 
252: 
253: \begin{figure*}[tb]
254: \begin{center}
255:  \includegraphics[width = 0.70\linewidth]{fig3.eps}
256: \end{center}
257: \caption{
258: (a) Standardized PDF for a single subject during constant
259: routine protocol [see Fig.~\ref{fig:fig1}(b)]. In solid lines,
260: we superimpose the PDF's using Castaing's equation. (b)
261: Dependence of the fitting parameter $\lambda^2$ of Castaing's
262: equation on the order of detrending polynomials. (c)
263: Dependence of the fitting parameter $\lambda^2$ on the scale $s$.
264: The error bars indicate the standard error of the group averages.
265: (d) Superposition of standardized PDF's at different scales
266: shown in Fig.~\ref{fig:fig2}a and \ref{fig:fig2}b, where we use
267: the scale range $8 \leq s \leq 2048$ and $8 \leq s \leq 4096$,
268: respectively. In solid lines, we superimpose the PDF using the
269: Castaing's equation ($\lambda^2 = 0.16$ for both groups).
270: }
271: \label{fig:fig3}
272: \end{figure*}
273: 
274: Struzik {\it et al.\/} \cite{Struzik04} recently demonstrated that
275: modifying the relative importance of either the sympathetic or
276: the parasympathetic branch of the autonomic nervous system leads
277: to a substantial decrease in $1/f$ scaling, showing that $1/f$
278: scaling in healthy heart rate requires the existence of and the
279: intricate balance between the antagonistic activity of these two
280: branches. They further suggest the view of cardiac
281: neuroregulation as a system in a critical state \cite{Bak87}, and
282: permanently out of equilibrium, in which concerted interplay of
283: the sympathetic and parasympathetic nervous systems is required
284: for preserving momentary `balance'. Our findings provide more
285: direct evidence for this. The precise mechanism
286: responsible for {\it critical\/} heart rate dynamics requires
287: further research.  It is of note, however, that there exists a 
288: physiological model for the dynamics of cardiac neuroregulation
289: \cite{Kotani02}, equipped with antagonistic and multiplicative
290: delayed feedback loops, within time scales where the critical
291: scale-invariance in heart rate is observed. 
292: The mechanism of the critical mode of operation could be clarified 
293: by investigating essential dynamics in such a `first principles', 
294: non-linear physiological model.
295: 
296: The functional advantage of the heart rate control system being in a
297: critical state remains an open question.  However, an analogy with
298: other critical phenomena might help to understand this.
299: Studies on transport properties through
300: complex networks \cite{Takayasu00,Valverdea02} have demonstrated
301: maximum efficiency of transportation at the critical point, which
302: is the  phase transition point from an `uncrowded' state to a
303: `congested' state in the transportation routes. % In the case of the
304: % heart, these would correspond typically to a state with central
305: % hypovolemia (uncrowded) and that with `congestive' heart failure
306: % (congested). Both cases are known to be associated with decreased
307: % cardiac output due to a lack of blood to pump out and difficulty
308: % in pumping out the blood, respectively \cite{Guyton}.
309: Thus, our results may indicate that the central neuroregulation continually
310: brings the heart to a critical state to maximize its functional
311: ability, coping with the continually changing pre- and after-load
312: on the heart. This may be particularly important in understanding
313: the widely reported evidence that decreased $1/f$ variability
314: \cite{1532,1705}, especially in the low frequency region \cite{1532,1305},
315: is associated with increased mortality in cardiac patients.
316: To date there have been no successful attempts to provide a
317: satisfactory explanation for this. We suggest for the first time
318: that a breakdown of the optimal control achieved in the critical
319: state might be related to this clinically important phenomenon.
320: 
321: This work was in part supported by grants from the Research Fellowships
322: of the Japan Society for the Promotion of Science for Young Scientists
323: (to K.K.) and from Japan Science and Technology Agency (to Y.Y.).
324: 
325: \begin{thebibliography}{28}
326: \medskip
327: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
328: \expandafter\ifx\csname bibnamefont\endcsname\relax
329:   \def\bibnamefont#1{#1}\fi
330: \expandafter\ifx\csname bibfnamefont\endcsname\relax
331:   \def\bibfnamefont#1{#1}\fi
332: \expandafter\ifx\csname citenamefont\endcsname\relax
333:   \def\citenamefont#1{#1}\fi
334: \expandafter\ifx\csname url\endcsname\relax
335:   \def\url#1{\texttt{#1}}\fi
336: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
337: \providecommand{\bibinfo}[2]{#2}
338: \providecommand{\eprint}[2][]{\url{#2}}
339: 
340: \bibitem[{\citenamefont{Peng et~al.}(1993)\citenamefont{Peng, Mietus,
341:   Hausdorff, Havlin, Stanley, and Goldberger}}]{1357}
342: \bibinfo{author}{\bibfnamefont{C.~K.} \bibnamefont{Peng}} {\it et al.},
343: %  \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Mietus}},
344: %  \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Hausdorff}},
345: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Havlin}},
346: %  \bibinfo{author}{\bibfnamefont{H.~E.} \bibnamefont{Stanley}},
347: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~L.}
348: %  \bibnamefont{Goldberger}},
349:   \bibinfo{journal}{Phys. Rev. Lett.}
350:   \textbf{\bibinfo{volume}{70}}, \bibinfo{pages}{1343} (\bibinfo{year}{1993}).
351: 
352: \bibitem[{\citenamefont{Yamamoto and Hughson}(1994)}]{1430}
353: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yamamoto}} \bibnamefont{and}
354:   \bibinfo{author}{\bibfnamefont{R.~L.} \bibnamefont{Hughson}},
355:   \bibinfo{journal}{Am. J. Physiol.} \textbf{\bibinfo{volume}{266}},
356:   \bibinfo{pages}{R40} (\bibinfo{year}{1994}).
357: 
358: \bibitem[{\citenamefont{Ivanov et~al.}(1999)\citenamefont{Ivanov, Amaral,
359:   Goldberger, Havlin, Rosenblum, Struzik, and Stanley}}]{1620}
360: \bibinfo{author}{\bibfnamefont{P.~C.} \bibnamefont{Ivanov}} {\it et al.},
361: %  \bibinfo{author}{\bibfnamefont{L.~A.~N.} \bibnamefont{Amaral}},
362: %  \bibinfo{author}{\bibfnamefont{A.~L.} \bibnamefont{Goldberger}},
363: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Havlin}},
364: %  \bibinfo{author}{\bibfnamefont{M.~G.} \bibnamefont{Rosenblum}},
365: %  \bibinfo{author}{\bibfnamefont{Z.~R.} \bibnamefont{Struzik}},
366: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.~E.}
367: %  \bibnamefont{Stanley}},
368:   \bibinfo{journal}{Nature}
369:   \textbf{\bibinfo{volume}{399}}, \bibinfo{pages}{461} (\bibinfo{year}{1999}).
370: 
371: \bibitem[{\citenamefont{Amaral et~al.}(2001)\citenamefont{Amaral, Ivanov,
372:   Aoyagi, Hidaka, Tomono, Goldberger, Stanley, and Yamamoto}}]{Amaral01}
373: \bibinfo{author}{\bibfnamefont{L.~A.~N.} \bibnamefont{Amaral}} {\it et al.},
374: %  \bibinfo{author}{\bibfnamefont{P.~C.} \bibnamefont{Ivanov}},
375: %  \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Aoyagi}},
376: %  \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Hidaka}},
377: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Tomono}},
378: %  \bibinfo{author}{\bibfnamefont{A.~L.} \bibnamefont{Goldberger}},
379: %  \bibinfo{author}{\bibfnamefont{H.~E.} \bibnamefont{Stanley}},
380: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yamamoto}},
381:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{86}},
382:   \bibinfo{pages}{6026} (\bibinfo{year}{2001}).
383: 
384: \bibitem[{\citenamefont{Frisch}(1995)}]{Frisch}
385: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Frisch}},
386:   \emph{\bibinfo{title}{Turbulence}} (\bibinfo{publisher}{Cambridge University
387:   Press}, \bibinfo{address}{Cambridge}, \bibinfo{year}{1995}),
388:   \bibinfo{edition}{1st} ed.
389: 
390: \bibitem[{\citenamefont{Lin and Hughson}(2001)}]{1759}
391: \bibinfo{author}{\bibfnamefont{D.~C.} \bibnamefont{Lin}} \bibnamefont{and}
392:   \bibinfo{author}{\bibfnamefont{R.~L.} \bibnamefont{Hughson}},
393:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{86}},
394:   \bibinfo{pages}{1650} (\bibinfo{year}{2001}).
395: 
396: \bibitem[{\citenamefont{Peng et~al.}(1994)\citenamefont{Peng, Buldyrev,
397:   Hausdorff, Havlin, Mietus, Simons, Stanley, and Goldberger}}]{1522}
398: \bibinfo{author}{\bibfnamefont{C.~K.} \bibnamefont{Peng}} {\it et al.},
399: %  \bibinfo{author}{\bibfnamefont{S.~V.} \bibnamefont{Buldyrev}},
400: %  \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Hausdorff}},
401: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Havlin}},
402: %  \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{Mietus}},
403: %  \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Simons}},
404: %  \bibinfo{author}{\bibfnamefont{H.~E.} \bibnamefont{Stanley}},
405: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~L.}
406: %  \bibnamefont{Goldberger}},
407:   \bibinfo{journal}{Integrative Physiol. Behav.
408:   Sci.} \textbf{\bibinfo{volume}{29}}, \bibinfo{pages}{283}
409:   (\bibinfo{year}{1994}).
410: 
411: \bibitem[{\citenamefont{Bigger et~al.}(1996)\citenamefont{Bigger, Steinman,
412:   Rolnitzky, Fleiss, Albrecht, and Cohen}}]{1532}
413: \bibinfo{author}{\bibfnamefont{J.~T.} \bibnamefont{Bigger},
414:   \bibfnamefont{Jr.}} {\it et al.},
415: % \bibinfo{author}{\bibfnamefont{R.~C.}
416: %  \bibnamefont{Steinman}}, \bibinfo{author}{\bibfnamefont{L.~M.}
417: %  \bibnamefont{Rolnitzky}}, \bibinfo{author}{\bibfnamefont{J.~L.}
418: %  \bibnamefont{Fleiss}},
419: %  \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Albrecht}}, \bibnamefont{and}
420: %  \bibinfo{author}{\bibfnamefont{R.~J.} \bibnamefont{Cohen}},
421:   \bibinfo{journal}{Circulation} \textbf{\bibinfo{volume}{93}},
422:   \bibinfo{pages}{2142} (\bibinfo{year}{1996}).
423: 
424: \bibitem[{\citenamefont{Huikuri et~al.}(2000)\citenamefont{Huikuri,
425:   M{\"{a}}kikallio, Peng, Goldberger, Hintze, and M{\o}ller}}]{1705}
426: \bibinfo{author}{\bibfnamefont{H.~V.} \bibnamefont{Huikuri}} {\it et al.},
427: %  \bibinfo{author}{\bibfnamefont{T.~H.} \bibnamefont{M{\"{a}}kikallio}},
428: %  \bibinfo{author}{\bibfnamefont{C.~K.} \bibnamefont{Peng}},
429: %  \bibinfo{author}{\bibfnamefont{A.~L.} \bibnamefont{Goldberger}},
430: %  \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Hintze}}, \bibnamefont{and}
431: %  \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{M{\o}ller}},
432:   \bibinfo{journal}{Circulation} \textbf{\bibinfo{volume}{101}},
433:   \bibinfo{pages}{47} (\bibinfo{year}{2000}).
434: 
435: \bibitem[{\citenamefont{Bigger et~al.}(1992)\citenamefont{Bigger, Fleiss,
436:   Steinman, Rolnitzky, Kleiger, and Rottman}}]{1305}
437: \bibinfo{author}{\bibfnamefont{J.~T.} \bibnamefont{Bigger},
438:   \bibfnamefont{Jr.}} {\it et al.},
439: % \bibinfo{author}{\bibfnamefont{J.~L.}
440: %  \bibnamefont{Fleiss}}, \bibinfo{author}{\bibfnamefont{R.~C.}
441: %  \bibnamefont{Steinman}}, \bibinfo{author}{\bibfnamefont{L.~M.}
442: %  \bibnamefont{Rolnitzky}}, \bibinfo{author}{\bibfnamefont{R.~E.}
443: %  \bibnamefont{Kleiger}}, \bibnamefont{and}
444: %  \bibinfo{author}{\bibfnamefont{J.~N.} \bibnamefont{Rottman}},
445:   \bibinfo{journal}{Circulation} \textbf{\bibinfo{volume}{85}},
446:   \bibinfo{pages}{164} (\bibinfo{year}{1992}).
447: 
448: \bibitem[{\citenamefont{Muzy et~al.}(1993)\citenamefont{Muzy, Bacry, and
449:   Arneodo}}]{Muzy93a}
450: \bibinfo{author}{\bibfnamefont{J.~F.} \bibnamefont{Muzy}},
451:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Bacry}}, \bibnamefont{and}
452:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Arneodo}},
453:   \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{47}},
454:   \bibinfo{pages}{875} (\bibinfo{year}{1993}).
455: 
456: \bibitem[{\citenamefont{Castaing et~al.}(1990)\citenamefont{Castaing, Gagne,
457:   and Hopfinger}}]{Castaing90}
458: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Castaing}},
459:   \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Gagne}}, \bibnamefont{and}
460:   \bibinfo{author}{\bibfnamefont{E.~J.} \bibnamefont{Hopfinger}},
461:   \bibinfo{journal}{Physica D} \textbf{\bibinfo{volume}{46}},
462:   \bibinfo{pages}{177} (\bibinfo{year}{1990}).
463: 
464: \bibitem[{\citenamefont{Hinrichsen et~al.}(2002)\citenamefont{Hinrichsen,
465:   Stenull, and Janssen}}]{Hinrichsen02}
466: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Hinrichsen}},
467:   \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Stenull}}, \bibnamefont{and}
468:   \bibinfo{author}{\bibfnamefont{H.-K.} \bibnamefont{Janssen}},
469:   \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{65}},
470:   \bibinfo{pages}{045104} (\bibinfo{year}{2002}).
471: 
472: \bibitem[{\citenamefont{Kadanoff et~al.}(1989)\citenamefont{Kadanoff, Nagel,
473:   Wu, and Zhou}}]{Kadanoff89}
474: \bibinfo{author}{\bibfnamefont{L.~P.} \bibnamefont{Kadanoff}} {\it et al.},
475: %  \bibinfo{author}{\bibfnamefont{S.~R.} \bibnamefont{Nagel}},
476: %  \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Wu}}, \bibnamefont{and}
477: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Zhou}},
478:   \bibinfo{journal}{Phys. Rev. A} \textbf{\bibinfo{volume}{39}},
479:   \bibinfo{pages}{6524} (\bibinfo{year}{1989}).
480: 
481: \bibitem[{\citenamefont{Stanley}(1999)}]{Stanley99}
482: \bibinfo{author}{\bibfnamefont{H.~E.} \bibnamefont{Stanley}},
483:   \bibinfo{journal}{Rev. Mod. Phys.} \textbf{\bibinfo{volume}{71}},
484:   \bibinfo{pages}{S358} (\bibinfo{year}{1999}).
485: 
486: \bibitem[{\citenamefont{Sethna et~al.}(2001)\citenamefont{Sethna, Dahmen, and
487:   Myers}}]{Sethna2001}
488: \bibinfo{author}{\bibfnamefont{J.~P.} \bibnamefont{Sethna}},
489:   \bibinfo{author}{\bibfnamefont{K.~A.} \bibnamefont{Dahmen}},
490:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{C.~R.} \bibnamefont{Myers}},
491:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{410}},
492:   \bibinfo{pages}{242} (\bibinfo{year}{2001}).
493: 
494: \bibitem[{\citenamefont{Sakata et~al.}(1999)\citenamefont{Sakata, Hayano,
495:   Mukai, Okada, and Fujinami}}]{Sakata99}
496: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Sakata}} {\it et al.},
497: %  \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Hayano}},
498: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Mukai}},
499: %  \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Okada}}, \bibnamefont{and}
500: %  \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Fujinami}},
501:   \bibinfo{journal}{Am. J. Physiol.} \textbf{\bibinfo{volume}{276}},
502:   \bibinfo{pages}{R1724} (\bibinfo{year}{1999}).
503: 
504: \bibitem[{\citenamefont{Aoyagi et~al.}(2003)\citenamefont{Aoyagi, Ohashi, and
505:   Yamamoto}}]{Aoyagi03}
506: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Aoyagi}},
507:   \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Ohashi}}, \bibnamefont{and}
508:   \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yamamoto}},
509:   \bibinfo{journal}{Am. J. Physiol.} \textbf{\bibinfo{volume}{285}},
510:   \bibinfo{pages}{R171} (\bibinfo{year}{2003}).
511: 
512: \bibitem[{\citenamefont{Schreiber and Schmitz}(2000)}]{Schreiber00}
513: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Schreiber}} \bibnamefont{and}
514:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Schmitz}},
515:   \bibinfo{journal}{Physica D} \textbf{\bibinfo{volume}{142}},
516:   \bibinfo{pages}{346} (\bibinfo{year}{2000}).
517: 
518: \bibitem[{\citenamefont{Ghashghaie et~al.}(1996)\citenamefont{Ghashghaie,
519:   Breymann, Peinke, Talkner, and Dodge}}]{Ghashghaie96}
520: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Ghashghaie}} {\it et al.},
521: %  \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Breymann}},
522: %  \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Peinke}},
523: %  \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Talkner}}, \bibnamefont{and}
524: %  \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Dodge}},
525:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{381}},
526:   \bibinfo{pages}{767} (\bibinfo{year}{1996}).
527: 
528: \bibitem[{\citenamefont{Mehta et~al.}(2002)\citenamefont{Mehta, Mills, Dahmen,
529:   and Sethna}}]{Mehta02}
530: \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Mehta}} {\it et al.},
531: %  \bibinfo{author}{\bibfnamefont{A.~C.} \bibnamefont{Mills}},
532: %  \bibinfo{author}{\bibfnamefont{K.~A.} \bibnamefont{Dahmen}},
533: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.~P.}
534: %  \bibnamefont{Sethna}},
535:   \bibinfo{journal}{Phys. Rev. E}
536:   \textbf{\bibinfo{volume}{65}}, \bibinfo{pages}{046139}
537:   (\bibinfo{year}{2002}).
538: 
539: \bibitem[{\citenamefont{Struzik et~al.}(2004)\citenamefont{Struzik, Hayano,
540:   Sakata, Kwak, and Yamamoto}}]{Struzik04}
541: \bibinfo{author}{\bibfnamefont{Z.~R.} \bibnamefont{Struzik}} {\it et al.},
542: %  \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Hayano}},
543: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Sakata}},
544: %  \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Kwak}}, \bibnamefont{and}
545: %  \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yamamoto}},
546:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{submitted}}
547:   (\bibinfo{year}{2004}).
548: 
549: \bibitem[{\citenamefont{Bak et~al.}(1987)\citenamefont{Bak, Tang, and
550:   Wiesenfeld}}]{Bak87}
551: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Bak}},
552:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Tang}}, \bibnamefont{and}
553:   \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Wiesenfeld}},
554:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{59}},
555:   \bibinfo{pages}{381} (\bibinfo{year}{1987}).
556: 
557: \bibitem[{\citenamefont{Kotani et~al.}(2002)\citenamefont{Kotani, Takamasu,
558:   Ashkenazy, Stanley, and Yamamoto}}]{Kotani02}
559: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Kotani}} {\it et al.},
560: %  \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Takamasu}},
561: %  \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Ashkenazy}},
562: %  \bibinfo{author}{\bibfnamefont{H.~E.} \bibnamefont{Stanley}},
563: %  \bibnamefont{and} \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yamamoto}},
564:   \bibinfo{journal}{Phys. Rev. E} \textbf{\bibinfo{volume}{65}},
565:   \bibinfo{pages}{051923} (\bibinfo{year}{2002}).
566: 
567: %\bibitem[{\citenamefont{de~Menezes M.~A. and Baraba\'si}(2004)}]{Menezes04}
568: %\bibinfo{author}{\bibnamefont{de~Menezes M.~A.}} \bibnamefont{and}
569: %  \bibinfo{author}{\bibfnamefont{A.-L.} \bibnamefont{Baraba\'si}},
570: %  \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{92}},
571: %  \bibinfo{pages}{028701} (\bibinfo{year}{2004}).
572: 
573: \bibitem[{\citenamefont{Takayasu et~al.}(2000)\citenamefont{Takayasu, Takayasu,
574:   and Fukuda}}]{Takayasu00}
575: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Takayasu}},
576:   \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Takayasu}}, \bibnamefont{and}
577:   \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Fukuda}},
578:   \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{277}},
579:   \bibinfo{pages}{248} (\bibinfo{year}{2000}).
580: 
581: \bibitem[{\citenamefont{Valverdea and Sol\'e}(2002)}]{Valverdea02}
582: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Valverdea}} \bibnamefont{and}
583:   \bibinfo{author}{\bibfnamefont{R.~V.} \bibnamefont{Sol\'e}},
584:   \bibinfo{journal}{Physica A} \textbf{\bibinfo{volume}{312}},
585:   \bibinfo{pages}{636} (\bibinfo{year}{2002}).
586: 
587: %\bibitem[{\citenamefont{Guyton and Hall}(2000)}]{Guyton}
588: %\bibinfo{author}{\bibfnamefont{A.~C.} \bibnamefont{Guyton}} \bibnamefont{and}
589: %  \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{Hall}},
590: %  \emph{\bibinfo{title}{Textbook of {M}edical {P}hysiology}}
591: %  (\bibinfo{publisher}{Elsevier Science}, \bibinfo{address}{Amsterdam},
592: %  \bibinfo{year}{2000}), \bibinfo{edition}{10th} ed.
593: 
594: \end{thebibliography}
595: 
596: \end{document}
597: 
598: 
599: 
600: