The healthy heartbeat is traditionally thought to be regulated according to the classical principle of homeostasis whereby physiologic systems operate to reduce variability and achieve an equilibrium‐like state [Physiol. Rev. 9, 399–431 (1929)]. However, recent studies [Phys. Rev. Lett. 70, 1343–1346 (1993); FractalsinBiologyandMedicine (Birkhauser‐Verlag, Basel, 1994), pp. 55–65] reveal that under normal conditions, beat‐to‐beat fluctuations in heart rate display the kind of long‐range correlations typically exhibited by dynamical systems far from equilibrium [Phys. Rev. Lett. 59, 381–384 (1987)]. In contrast, heart rate time series from patients with severe congestive heart failure show a breakdown of this long‐range correlation behavior. We describe a new method—detrended fluctuation analysis (DFA)—for quantifying this correlation property in non‐stationary physiological time series. Application of this technique shows evidence for a crossover phenomenon associated with a change in short and long‐range scaling exponents. This method may be of use in distinguishing healthy from pathologic data sets based on differences in these scaling properties.

1.
R. I. Kitney and O. Rompelman, The Study of Heart-Rate Variability (Oxford University Press, Oxford, 1980).
2.
A. L.
Goldberger
,
D. R.
Rigney
, and
B. J.
West
,
Sci. Am.
262
,
42
49
(
1990
).
3.
C.-K.
Peng
,
J.
Mietus
,
J. M.
Hausdorff
,
S.
Havlin
,
H. E.
Stanley
, and
A. L.
Goldberger
,
Phys. Rev. Lett.
70
,
1343
1346
(
1993
).
4.
C.-K. Peng, S. V. Buldyrev, J. M. Hausdorff, S. Havlin, J. E. Mietus, M. Simons, H. E. Stanley, and A. L. Goldberger, in Fractals in Biology and Medicine, edited by T. F. Nonnenmacher, G. A. Losa, and E. R, Weibel (Birkhaüser-Verlag, Basel, 1994), pp. 55–65.
5.
P.
Bak
,
C.
Tang
, and
K.
Wiesenfeld
,
Phys. Rev. Lett.
59
,
381
384
(
1987
).
6.
H. E. Stanley, Introduction to Phase Transitions and Critical Phenomena (Oxford University Press, London, 1971).
7.
M. N.
Levy
,
Circ. Res.
29
,
437
445
(
1971
).
8.
C.-K.
Peng
,
S. V.
Buldyrev
,
S.
Havlin
,
M.
Simons
,
H. E.
Stanley
, and
A. L.
Goldberger
,
Phys. Rev. E
49
,
1691
1695
(
1994
).
9.
Computer software of DFA algorithm is available upon request; contact C.-K. Peng (e-mail: peng[chaos.bih.harvard.edu).
10.
S. V.
Buldyrev
,
A. L.
Goldberger
,
S.
Havlin
,
C.-K.
Peng
,
H. E.
Stanley
, and
M.
Simons
,
Biophys. J.
65
,
2675
2681
(
1993
).
11.
S. M.
Ossadnik
,
S. V.
Buldyrev
,
A. L.
Goldberger
,
S.
Havlin
,
R. N.
Man- tegna
,
C.-K.
Peng
,
M.
Simons
, and
H. E.
Stanley
,
Biophys. J.
67
,
64
70
(
1994
).
12.
J. M.
Hausdorff
,
C.-K.
Peng
,
Z.
Ladin
,
J. Y.
Wei
, and
A. L.
Goldberger
,
J. Appl. Physiol.
78
,
349
358
(
1995
).
13.
E. W. Montroll and M. F. Shlesinger, in Nonequilibrium Phenomena II. From Stochastics to Hydrodynamics, edited by J. L. Lebowitz and E. W. Montroll (North-Holland, Amsterdam, 1984), pp 1–121.
14.
S.
Havlin
,
R. B.
Selinger
,
M.
Schwartz
,
H. E.
Stanley
, and
A.
Bunde
,
Phys. Rev. Lett.
61
,
1438
1441
(
1988
).
15.
W. H.
Press
,
Comments Astrophys.
7
,
103
119
(
1978
).
16.
C.-K.
Peng
,
S.
Buldyrev
,
A. L.
Goldberger
,
S.
Havlin
,
F.
Sciortino
,
M.
Simons
, and
H. E.
Stanley
,
Nature
356
,
168
170
(
1992
).
17.
S. V. Buldyrev, A. L. Goldberger, S. Havlin, C.-K. Peng, and H. E. Stanley, in Fractals in Science, edited by A. Bunde and S. Havlin (Springer-Verlag, Berlin, 1994), pp. 48–87.
18.
ECG recordings of Holter monitor tapes were processed both manually and in a fully automated manner using our computerized beat recognition algorithm (Aristotle). Abnormal beats were deleted from each data set. The deletion has practically no effect on the DFA analysis since less than \% of total beats were removed. Patients in the heart failure group were receiving conventional medical therapy prior to receiving an investigational cardiotonic drug; see
D. S.
Baim
et al.,
J. Am. Coll. Cardiol.
7
,
661
670
(
1986
).
19.
A. L.
Goldberger
,
D. R.
Rigney
,
J.
Mietus
,
E. M.
Antman
, and
S.
Greenwald
,
Experientia
44
,
983
987
(
1988
).
20.
Typical regression fit shows excellent linearity of double log graph (indicated by correlation coefficient r>0.97) for both groups. However, usually data from healthy subjects show even better linearity on log-log plots than data from subjects with heart disease. Our estimate of a is consistent with the previous analysis in Ref. 3. Note, however, that in Ref. 3 the analysis was performed on the interbeat increment data to avoid the problem of non-stationarity. Therefore, the scaling exponent computed in Ref. 3 is smaller than the a exponent computed by DFA by a value of 1 (due to the integration process in DFA).
21.
A. L.
Goldberger
and
B. J.
West
,
Yale J. Biol. Med.
60
,
421
435
(
1987
).
22.
M.
Kobayashi
and
T.
Musha
,
IEEE Trans. Biomed. Eng.
BE-29
,
456
457
(
1982
).
23.
W. B.
Cannon
,
Physiol. Rev.
9
,
399
431
(
1929
).
24.
A further refinement (not presented here) may be obtained by not arbitrarily setting the crossover scale to be 16 beats for all data sets. Instead, each individual data set could have its own ranges for fitting α1 and α2 that depend on the specific crossover point in the given data set.
25.
C.-K.
Peng
,
S. V.
Buldyrev
,
A. L.
Goldberger
,
S.
Havlin
,
M.
Simons
, and
H. E.
Stanley
,
Phys. Rev. E
47
,
3730
3733
(
1993
).
26.
We also tested these calculations by varying the fitting range for α2. We find that the results are very similar when we measure α2 from 16 beats to 128 beats. However, when we move the upper fitting range for α2 from 128 beats to 256 beats or more, the pathologic data sets show larger variation of α2 leading to less obvious separation from normal subjects. This is partly due to the fact that, for finite length data sets, the calculation error of F(n) increases with n.25 Therefore, scaling exponents obtained over larger values of n will have greater uncertainty.
This content is only available via PDF.
You do not currently have access to this content.