Professional Documents
Culture Documents
Ferrario 2006
Ferrario 2006
Ferrario 2006
Abstract—This paper considers the multiscale entropy (MSE) A first goal of this paper is to examine the behavior of MSE
approach for estimating the regularity of time series at different for simulated signals, in order to understand the informative
scales. Sample entropy (SampEn) and approximate entropy content that can be obtained from the signal itself. Moreover,
(ApEn) are evaluated in MSE analysis on simulated data to
enhance the main features of both estimators. we investigated the sensitivity of MSE values with respect to
We applied the approximate entropy and the sample entropy the time series length.
estimators to fetal heart rate signals on both single and multiple A second goal is to confirm the usefulness of entropy estima-
scales for an early identification of fetal sufferance antepartum. tors on a real diagnostic problem, consisting of the dichotomic
Our results show that the ApEn index significantly distinguishes classification of normal and suffering fetuses during pregnancy.
suffering from normal fetuses between the 30th and the 35th week
of gestation. Furthermore, our data shows that the MSE entropy In order to achieve this second goal, we applied these regularity
values are reliable indicators of the fetal distress associated with estimators to fetal heart rate (FHR) time series recorded during
the presence of a pathological condition at birth. cardiotocographic (CTG) monitoring antepartum.
Index Terms—Cardiotocography, fetal heart rate, fetal patholo- Fetal distress (FD) can occur for several reasons and it
gies, nonlinear parameters, spectral analysis. can appear during various stages of pregnancy with different
severity degrees and symptoms. In general, FD is characterized
by the reduction of the maternal-fetal respiratory exchanges,
I. INTRODUCTION which mainly causes a reduction of fetal blood oxygenation
and heap of carbon dioxide. It induces a state of anaerobic
I N the last years, many studies tried to demonstrate the non-
linear nature of the heart rate variability (HRV) signal. These
attempts encountered many practical problems, such as the lim-
metabolism, which is associated with an increase of the acid
metabolite concentration [5].
ited length of the time series introducing approximations of The early identification of a distress status is among the fun-
theoretical methods and consequently a limitation of purposes. damental goals of fetal monitoring antepartum and it becomes
An example is the approximate entropy (ApEn) introduced by more and more important as the advances in neonatal pathology
Pincus [1], [2], which provides a measure of signal regularity. allow newborns surviving at earlier stages of premature deliv-
Differently from the original goal of the Kolmogorov entropy, eries. As a matter of fact, the early detection of a life-threat-
which aimed at investigating the dynamics of the generating ening condition, such as a severe distress, helps the obstetrician
system, eventually up to confirm its chaotic nature, the approx- to decide for an immediate action (e.g., to induce delivery by ce-
imate entropy (ApEn) index was conceived to characterize the sarean section). Furthermore, a correct identification of the fetal
time series without any hypothesis on the nature of the gener- state allows avoiding both false positives (unnecessary cesarean
ating system. This last approach demonstrated its practical use- sections) and false negatives (pregnancy prosecution in case of
fulness in many experimental conditions especially when the severe sufferance).
identification of a precise model structure was impossible or in- Nowadays, in most clinical situations, the obstetrician must
appropriate. rely on his/her ability to evaluate the CTG tracings by eye in-
Recently, Costa et al. [3], [4] made a further step toward the spection or with the help of some classical morphological in-
characterization of the signal structure by itself, by introducing dexes of proven diagnostic weakness [6]. Thus, quantitative re-
the multiscale entropy (MSE) measurement, which is an indi- liable clinical tools are required for early detecting the appear-
cator of the regularity of the signal at different time scales. ance of fetal sufferance. Our final goal is then to investigate
whether some regularity estimators, applied to the FHR signal,
can provide quantitative indexes revealing the presence of FD
Manuscript received October 12, 2004, revised May 30, 2005. Asterisk indi- antepartum.
cates corresponding author.
M. Ferrario and S. Cerutti are with the Dipartimento di Bioingegneria, Uni-
versity Politecnico di Milano, 20133 Milano, Italy. II. METHODS
*M. G. Signorini is with the Dipartimento di Bioingegneria, piazza Leonardo
da Vinci 32, University Politecnico di Milano, 20133 Milano, Italy (e-mail: The development of chaos theory in the last decades sup-
mariagabriella.signorini@polimi.it). plied the framework to study nonlinear dynamics through a new
G. Magenes is with the Dipartimento di Informatica e Sistemistica, University
of Pavia, 27100 Pavia, Italy (e-mail: giovanni@bioing.unipv.it). approach. Many efforts were made to explain the complex be-
Digital Object Identifier 10.1109/TBME.2005.859809 havior of deterministic systems with the presence of nonlinear
0018-9294/$20.00 © 2006 IEEE
120 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 1, JANUARY 2006
features instead of the usual stochastic models. The most direct As Pincus himself noticed [1], [2], [12], approximate entropy
link between chaos theory and the real world is the analysis of is affected by a bias effect. In addition, ApEn strongly
time series from real systems in terms of nonlinear dynamics [7]. depends on the record length. The correction of the bias effect
In some situations, when the a priori knowledge about the gen- is not trivial since the straightforward removal of self-counting
erating system is deep enough, it is possible to precisely charac- could origin a high sensitivity to outliers. Furthermore, Richman
terize the structure of the system itself from the observed time and Moorman [13] demonstrated that in some cases ApEn lacks
series. In these cases, the study involves a mathematical charac- of consistency. Its computation in irregular time series is af-
terization of the phase space or the estimation of some descrip- fected by a bias, which may cause an overestimate of the en-
tors of the attractor. This approach leads to the reconstruction of tropy value, mostly when values are very small.
the attractor in an embedding space, estimating some invariants However, the application of ApEn statistics to real data often
such as fractal dimension and Lyapunov exponents [8]–[11]. obtains good results, giving general information about the reg-
Nevertheless, it is often impossible to obtain sufficient in- ularity and the persistence of the signal. This fact has led to an
sight on biological systems from signals recorded by noninva- extensive use of ApEn in clinical applications, e.g., the analysis
sive techniques. The knowledge we have on their behavior is of inter-beat RR signals in the study of cardiovascular diseases
limited by their intrinsic complexity, which is the result of nu- [14]–[16].
merous interacting mechanisms contributing to the physiolog- Entropy methods exploit a symbolic representation of the
ical performance. The concept of complexity does not corre- time series. Despite the severe reduction of information, they are
spond to a unique and rigorous definition. Sometimes it is re- able to enhance relevant features of the signal. ApEn, and more
lated to the difficulties we have in describing or understanding generally coarse-grained entropies, can be useful to track quali-
a signal. tative changes in time series patterns, without precisely charac-
Pincus [1] observed that the classic Kolmogorov-Sinai en- terizing the generating system.
tropy assumes infinite value for processes with superimposed Richman and Moorman [13], [17] developed a modification
stochastic noise. Thus, it is unable to distinguish processes that of the ApEn algorithm, named sample entropy (SampEn), in
differ in complexity. In this case, the word complexity refers order to remove some of the deficiencies reported here.
to the unpredictability of the system state location, once the ini- The differences with respect to ApEn are: 1) self-matches are
tial conditions are known. The less predictable the states are, the not counted; 2) only the first N- vectors of length are con-
more complex the system is. The association of unpredictability sidered; and 3) the conditional probabilities are not estimated in
with complexity leads to the conclusion that Gaussian noise is a template manner. Thus, the probability measure is computed
very complex because all points in the phase space are a prob- directly as the logarithm of conditional probability instead of
able state of the system and the state location is unpredictable. the ratio of the logarithmic sums.
In contrast, a periodic dynamical system with period 2 owns an SampEn shows a relative consistency in cases where ApEn
extremely low complexity level: in fact, all trajectories meet two does not. Furthermore the values of SampEn agree with the the-
points with probability 1. Nevertheless most biological systems oretical values expected for a uniform random noise time series
lie between the two extremes of this complexity scale. much more than the ApEn values, even for very short time se-
Starting from these considerations, Pincus proposed a family ries [13].
of statistics, called Approximate Entropy [2], aimed at mea- Both ApEn and SampEn supply a single index concerning
suring the signal regularity i.e., the presence of similar patterns the general behavior of the time series, but they cannot reveal
in the time series. the underlying dynamics of the generating system. Thus, if the
Given N data points ; 1 , the algorithm signal X has a lower entropy value than the signal Y, we can
constructs sequences and it computes, for each only say that X is more regular than Y. If the original purpose
1, the quantity of entropy was to identify chaotic dynamics, the statistics ApEn
and SampEn have changed the perspective by providing a figure
number of
related to the regularity of the time series at the original time
(1) scale.
measures, with a tolerance , the regularity of patterns Several methods have been proposed in the literature, for
by comparing them to a given pattern of length ( and are achieving a thorough analysis at different scales. Among them
fixed values: is the detail level at which the signal is analyzed, wavelet-transform modulus-maxima method (WTMM) appears
is a threshold, which filters out irregularities). a very promising tool [18]. Other methods, calculating the Hurst
The regularity parameter is defined as exponent estimate long range dependence scaling exponent and
statistical self-similarity as well as the presence of long range
ApEn correlation and “memory” in the process.
Costa et al. [3], [4] introduced the so-called MSE method.
where The procedure considers a time series of N points and con-
structs consecutive coarse-grained time series , as a func-
1 tion of the factor
1
ApEn is the estimator of this 1 (2)
parameter for an experimental time series of fixed length N.
FERRARIO et al.: COMPARISON OF ENTROPY-BASED REGULARITY ESTIMATORS 121
III. SIMULATIONS
To investigate the information provided by the MSE method
regarding signal characterization, we evaluated the behavior
of MSE by simulating a set of known signals. We selected an
example of stochastic and completely irregular signal (white
noise), a correlated and multiscale signal (1/f noise), a corrupted
deterministic signal (MIX process) and a chaotic deterministic 6
Fig. 1. Results of the MSE analysis (avg. std.) for simulated signals of 1/f
signal (logistic map). We chose this standard set of signals to noise. Each symbols refers to a samples number N of analyzed time series. All
generated signals have the same mean and std.
allow a comparison with the original works by Costa et al. [3],
[4], [19].
At first we considered which measure of entropy is suitable
for a multiscale analysis. We used both the sample entropy and
the ApEn as entropy measure for the multiscale approach, com-
paring their performance.
The adopted parameters were 2, 0.15 (SD:
standard deviation of the original time series), as reported in the
original works by Costa et al. [3], [4]. However, the MSE anal-
ysis, reported in the cited works, considers series collected from
24-h Holter recordings, noticeably longer than our 1-h FHR
tracings.
For this reason we applied the MSE to signals of different
length N ( 1000, 5000, 7000, 10 000, 15 000, and 30 000
samples) and we studied the index dependence on the time se-
ries length. Fig. 1 shows the results of the simulations in time
series from a 1/f noise process. Series were built by generating a
1/f power spectral density, with random phases (uniformly dis-
tributed on [ 2 , 2 ]) and by applying the inverse fast Fourier 6
Fig. 2. Results of the MSE analysis (avg. std.) for simulated signals, number
transform. Entropy measures, plotted for 1/f noise, show persis- of points N = 7000. The higher values refer to random noise, the lower ones
tent values for all scale factors, with the exception of the shortest
1 1 0
refer to sequences generated by the logistic map x + 1 = 4 x (1 x ).
TABLE I
CTG RECORDINGS SELECTED FOR THE ANALYSIS
TABLE II
STANDARD PARAMETERS FOR THE FETAL SURVEILLANCE IN NORMAL AND SUFFERING FETUSES
* =p < y
0.05 for the Student test, rank sum test, and Kolmogorov–Smirnov test (normal versus suffering). = symbol means statistical significance both for
0 < 0 <
ANOVA and Kruskall–Wallis test (p value 0.01) and Scheffè test (p value 0.05). N = Normal; RS = RecSuf; NRS = NotRecSuf.
(multiscale) time windows are able to significantly discriminate to a prevalence of activity periods in healthy fetuses analyzed at
the true distressed and pathological fetuses from the healthy the same gestational age.
and not suffering ones, although MSE analysis and single At this stage of gestation (30–35 weeks), in suffering fetuses,
ApEn/SampEn index apparently seem to provide contradictory the neural regulatory mechanisms do not yet show a clear al-
results. teration but the prevalence of quiet periods in the FHR could
The MSE values of normal fetuses increase, as the time se- be a marker of increasing distress conditions which could lead,
ries is progressively coarse grained. In this case, values ob- during the ongoing pregnancy, to pathology of the fetus.
tained in distressed fetuses show that their HRV signal has a Longer quiet periods might mean that the disease is starting
different signal structure (more regular, with a nearly sinusoidal but it does not yet affect the neural regulatory mechanisms
behavior) than normal FHR, which reveals an increasing trend controlling the heart frequency. The consequence is an initial
that is distinctive of more irregular signal patterns. Comparison loss of awakeness of the fetus (higher entropy) before a clear
with results of MIX(p) processes reinforces this hypothesis. pathology emerges, changing the behavior of regulatory mech-
The MSE approach for the whole recording time (7200 anisms (lower entropy, as it happens in the class of fetuses with
points) shows that pathological fetuses present lower entropy pathology at birth).
values than healthy ones at all scale factors, supporting the idea In conclusion, entropy indexes show their ability in inves-
that, as in adults, pathological conditions induce an increase of tigating the structure of biological time series. FHR analysis
signal regularity in the long-term, due to a possible loss of com- points out that the entropy estimators are sensitive to short-term
plexity (information reduction) in the regulatory mechanisms. and long-term signal patterns. In the short term, they catch local
The results of the MSE approach show that FHR variability irregularities as the dominant mechanism. The long term anal-
possesses a scale organization, which can be only highlighted ysis draws advantage from the severe reduction of information
by considering the entire signal. In fact, we observe that the which is imposed by the scaling procedure to enhance relevant
MSE SampEn at scale 1 (considering all the 7200 points) is features in the FHR signal. Results evidence a quite marked de-
lower than we previously obtained over 360 points. tection of severely distressed fetuses on the basis of differences
in the variability signal structure.
The results of both entropies (ApEn and SampEn), computed
These considerations reinforce the interest toward the use of
as a single averaged index over short time windows (360 points),
advanced and nonconventional signal processing methods in the
show an opposite behavior with respect to the multiscale ap-
clinical routine. The discriminating ability of regularity mea-
proach. Their interpretation is not trivial, and can be found by
sures enhances the emergence of information with biological
considering both the mathematical characteristics of entropy es-
importance from the time series structure.
timators over short time sequences and the development of FD
during pregnancy.
ApEn and SampEn resulting from short FHR time series in ACKNOWLEDGMENT
the original time scale are more sensitive to the local irregular- The authors would like to thank Prof. D. Arduini for his sup-
ities (local differences) of signal patterns. In fact, as it could be port in the data collection process at the Ob-Gyn Clinic, Hospital
expected, both ApEn and SampEn show higher values than in “Fatebenefratelli,” Rome, Italy, and Dr. S. Ramat for revising
the multiscale approach. As a matter of fact, the multiscale anal- our paper.
ysis over the entire signal does not enhance the abrupt events,
which are filtered out by the general trend of the signal for all REFERENCES
scale factors (even for factor 1, thus, the same scale used in short
[1] S. M. Pincus, “Approximate entropy (ApEn) as complexity measure,”
time series). The ApEn or SampEn values computed over short Chaos, vol. 5, no. 1, pp. 110–117, Mar. 1995.
windows provide instead the “local” information content corre- [2] S. M. Pincus, I. M. Gladstone, and R. A. Ehrenkranz, “A regularity
lated to the events. statistic for medical data analysis,” J. Clin. Monit., vol. 7, no. Oct., pp.
335–345, 1991.
The inversion of entropy levels in a short time analysis should [3] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy to
not be surprising. Sometimes it happens that both ApEn and distinguish physiologic and synthetic RR time series,” Comput. Cardiol.,
SampEn algorithms, when applied to some pathologic time se- vol. 29, pp. 137–140, 2002.
[4] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy anal-
ries, provide higher entropy values than those obtained from free ysis of complex physiologic time series,” Phys. Rev. Lett., vol. 89, no. 6,
running physiologic systems under healthy condition [19]. p. 068 102, Aug. 2002.
[5] H. M. McNamara and G. A. Dildy, “Continuous intrapartum pH, pCO2,
In relation to our FD problem, a possible interpretation of this SpO2 monitoring,” Obstet. Gynecol. Clin. N. Am., vol. 26, pp. 671–693,
experimental evidence can be found by comparing the obtained Dec. 1999.
results with those reported in a previous work [24] related to [6] P. Malcus, “Antenatal fetal surveillance,” Curr. Opin. Obstet. Gynaecol.,
vol. 16, pp. 123–128, 2004.
normal fetuses submitted to vibroacoustic stimulation. In that [7] H. Kantz and T. Schreiber, Nonlinear Time Series Analysis, 2 ed. Cam-
work, when the fetus was in a quiet period before the stimulus bridge, U.K.: Cambridge Univ. Press, 2004.
was given, the ApEn values were higher than in the activity pe- [8] T. S. Parker and L. O. Chua, “Chaos: a tutorial for engineers,” Proc.
IEEE, vol. 75, no. 8, pp. 982–1008, Aug. 1987.
riod after the vibroacoustic stimulation. We can, thus, make the [9] E. Ott, Chaos in Dynamical Systems. New York: Cambridge Univ.
reasonable hypothesis that higher entropy values found in suf- Press, 1993, pp. 1–22.
fering fetuses by averaging ApEn values, computed on succes- [10] J. P. Eckmann and D. Ruelle, “Ergodic theory of chaos and strange at-
tractors,” Rev. Modern Phys., vol. 57, pp. 617–654, Jul. 1985.
sive 3-min windows for the whole 1-h recording, can be due to [11] P. Grassberger and I. Procaccia, “Estimation of the Kolmogorov entropy
a prevalence of quiet periods in suffering fetuses, with respect from a chaotic signal,” Phys. Rev. A, vol. 28, pp. 2591–2593, May 1983.
FERRARIO et al.: COMPARISON OF ENTROPY-BASED REGULARITY ESTIMATORS 125
[12] S. M. Pincus, “Approximate entropy as a measure of system complexity,” Maria G. Signorini received the Ph.D. degree in
in Proc. Nat. Acad. Sci. USA, vol. 88, Mar. 1991, pp. 2297–2301. 1995 from the Polytechnic University, Milano, Italy.
[13] J. S. Richman and J. R. Moorman, “Physiological time-series analysis She is currently an Associate Professor in the
using approximate entropy and sample entropy,” Am. J. Physiol. Heart Biomedical Engineering Department at the same
Circ. Physiol., vol. 278, pp. H2039–2049, Jun. 2000. University. Her main research interests are nonlinear
[14] T. H. Mäkikallio et al., “Dynamic analysis of heart rate may predict analysis and modeling of biological signals and
subsequent ventricular tachycardia after myocardial infarction,” Am. J. systems regarding, in particular, the cardiovascular
Cardiol., vol. 80, no. 6, pp. 779–783, Sep. 1997. system dynamics. She cooperates in industrial
[15] K. K. L. Ho et al., “Predicting survival in heart failure case and control projects (HP Italy, Sister-Fresenius Medical Care)
subjects by use of fully automated methods for deriving nonlinear and on biomedical signal processing and instrumentation
conventional indexes of heart rate dynamics,” Circulation, vol. 96, no. improvement.
3, pp. 842–848, 1997.
[16] H. Fukuta et al., “Prognostic value of nonlinear heart rate dynamics in
hemodialysis patients with coronary artery disease,” Kidney Int., vol. 64,
no. 2, pp. 641–648, 2003.
[17] D. E.Lake , J. S. Richman, and J. R. Moorman, “Sample entropy analysis Giovanni Magenes received the “Laurea” degree
of neonatal heart rate variability,” Am. J. Physiol. Regul. Integr. Comp. in electronic engineering from the University of
Physiol., vol. 283, pp. R789–R797, Sep. 2002. Pavia, Pavia, Italy, in 1981 and the Ph.D. degree
[18] B. Audry, E. Bacry, J.-F. Muzy, and A. Arneodo, “Wavelet-based esti- in bioengineering from the Polytechnic of Milano,
mators of scaling behavior,” IEEE Trans. Inform. Theory, vol. 48, no. Milano, Italy, in 1986.
11, pp. 2938–2954, Nov. 2002. Since 1988, he is with the Dipartimento di
[19] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy anal- Informatica e Sistemistica, University of Pavia,
ysis of biological signals,” Phys. Rev. E, vol. 71, p. 021 906, 2005. where he is now full Professor of Biomedical Signal
[20] R. Mantel, H. P. van Geijn, F. J. Caron, J. M. Swartjes, E. E. van Wo- Processing. His main interests are in intelligent
erden, and H. W. Jongsma, “Computer analysis of antepartum fetal heart biosignal analysis, neural networks applications in
rate: 1. Baseline determination,” Int. J. Biomed. Comput., vol. 25, no. 4, robotics and biomedicine, and sensorimotor control
pp. 261–272, 1990. in men and machines.
[21] G. Magenes, M. G. Signorini., D. Arduini, F. Lunghi, and A. Bonalumi. Dr. Magenes is member of the European Neural Networks Society, and the
Development of a new system for cardiotocographic automatic analysis. Barany Society.
presented at Proc. 22nd Int. Conf. IEEE-EMBS Soc.. [CD-ROM]
[22] M. G. Signorini, G. Magenes, S. Cerutti, and D. Arduini, “Linear and
nonlinear parameters for the analysis of fetal heart rate signal from car-
diotocographic recordings,” IEEE Trans Biomed. Eng., vol. 50, no. 3,
pp. 365–375, Mar. 2003. Sergio Cerutti (M’81–SM’97–F’02) is Professor
[23] H. P.Van-Geijn , “Developments in CTG analysis,” Baillieres Clin. Ob- of Biomedical Signal and Data Processing at the
stet. Gynaecol., vol. 10, no. 2, pp. 185–209, 1996. Department of Biomedical Engineering of the
[24] G. Magenes, M. G. Signorini, D. Arduini, and S. Cerutti, “Fetal Heart Polytechnic University of Milan, Milan, Italy.
rate variability due to vibroacoustic stimulation: linear and nonlinear Since November 2000, he is the Chairman of the
contribution,” Meth. Inf. Med., vol. 43, no. 1, pp. 47–51, 2004. department. His research interests are mainly in
the following topics: biomedical signal processing
(ECG, blood pressure signal and respiration, car-
diovascular variability signals, EEG, and evoked
Manuela Ferrario received the Italian degree potentials), cardiovascular modeling, neurosciences,
(Laurea) from the Polytechnic University, Milano, medical informatics, and regulation and standards
Italy, in 2003. She received a scholarship for the in medical equipments and devices. Since March 1983, he has also taught a
development of a prototype clinical decision support course at a graduate level on Biomedical Signal Processing at Engineering
system and the improvement of the current decision Faculties (Milan and Rome) and at Specialization Schools of Medical Faculties
scores for a correct and early evaluation of fetal (Milan and Roma).
distress. She is currently working towards the Ph.D. Prof. Cerutti was an elected member of IEEE-EMBS AdCom (Region 8)
degree in the Biomedical Engineering Department for the period 1993–1996. He is an Associate Editor of IEEE TRANSACTIONS
at the same university. BIOMEDICAL ENGINEERING. He is a member of the Steering Committee of
Her main research interests are nonlinear analysis, the IEEE-EMBS Summer School on Biomedical Signal Processing; he was
fetal heart rate analysis, and multiscale and fractal the local organizer of three IEEE-EMBS Summer Schools held in Siena in
analysis. 1995/1999/2003.