IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO.
1, JANUARY 2006
Comparison of Entropy-Based Regularity Estimators:
Application to the Fetal Heart Rate Signal for the
Identification of Fetal Distress
Manuela Ferrario, Maria G. Signorini*, Giovanni Magenes, and Sergio Cerutti, Fellow, IEEE
Abstract—This paper considers the multiscale entropy (MSE)
approach for estimating the regularity of time series at different
scales. Sample entropy (SampEn) and approximate entropy
(ApEn) are evaluated in MSE analysis on simulated data to
enhance the main features of both estimators.
We applied the approximate entropy and the sample entropy
estimators to fetal heart rate signals on both single and multiple
scales for an early identification of fetal sufferance antepartum.
Our results show that the ApEn index significantly distinguishes
suffering from normal fetuses between the 30th and the 35th week
of gestation. Furthermore, our data shows that the MSE entropy
values are reliable indicators of the fetal distress associated with
the presence of a pathological condition at birth.
Index Terms—Cardiotocography, fetal heart rate, fetal patholo-
gies, nonlinear parameters, spectral analysis.
I. INTRODUCTION
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-
ited length of the time series introducing approximations of
theoretical methods and consequently a limitation of purposes.
An example is the approximate entropy (ApEn) introduced by
Pincus [1], [2], which provides a measure of signal regularity.
Differently from the original goal of the Kolmogorov entropy,
which aimed at investigating the dynamics of the generating
system, eventually up to confirm its chaotic nature, the approx-
imate entropy (ApEn) index was conceived to characterize the
time series without any hypothesis on the nature of the gener-
ating system. This last approach demonstrated its practical use-
fulness in many experimental conditions especially when the
identification of a precise model structure was impossible or in-
appropriate.
Recently, Costa et al. [3], [4] made a further step toward the
characterization of the signal structure by itself, by introducing
the multiscale entropy (MSE) measurement, which is an indi-
cator of the regularity of the signal at different time scales.
Manuscript received October 12, 2004, revised May 30, 2005. Asterisk indi-
cates corresponding author.
M. Ferrario and S. Cerutti are with the Dipartimento di Bioingegneria, Uni-
versity Politecnico di Milano, 20133 Milano, Italy.
*M. G. Signorini is with the Dipartimento di Bioingegneria, piazza Leonardo
da Vinci 32, University Politecnico di Milano, 20133 Milano, Italy (e-mail:
mariagabriella.signorini@polimi.it).
G. Magenes is with the Dipartimento di Informatica e Sistemistica, University
of Pavia, 27100 Pavia, Italy (e-mail: giovanni@bioing.unipv.it).
Digital Object Identifier 10.1109/TBME.2005.859809
0018-9294/$20.00 © 2006 IEEE
119
A first goal of this paper is to examine the behavior of MSE
for simulated signals, in order to understand the informative
content that can be obtained from the signal itself. Moreover,
we investigated the sensitivity of MSE values with respect to
the time series length.
A second goal is to confirm the usefulness of entropy estima-
tors on a real diagnostic problem, consisting of the dichotomic
classification of normal and suffering fetuses during pregnancy.
In order to achieve this second goal, we applied these regularity
estimators to fetal heart rate (FHR) time series recorded during
cardiotocographic (CTG) monitoring antepartum.
Fetal distress (FD) can occur for several reasons and it
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,
which mainly causes a reduction of fetal blood oxygenation
and heap of carbon dioxide. It induces a state of anaerobic
metabolism, which is associated with an increase of the acid
metabolite concentration [5].
The early identification of a distress status is among the fun-
damental goals of fetal monitoring antepartum and it becomes
more and more important as the advances in neonatal pathology
allow newborns surviving at earlier stages of premature deliv-
eries. As a matter of fact, the early detection of a life-threat-
ening condition, such as a severe distress, helps the obstetrician
to decide for an immediate action (e.g., to induce delivery by ce-
sarean section). Furthermore, a correct identification of the fetal
state allows avoiding both false positives (unnecessary cesarean
sections) and false negatives (pregnancy prosecution in case of
severe sufferance).
Nowadays, in most clinical situations, the obstetrician must
rely on his/her ability to evaluate the CTG tracings by eye in-
spection or with the help of some classical morphological in-
dexes of proven diagnostic weakness [6]. Thus, quantitative re-
liable clinical tools are required for early detecting the appear-
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
antepartum.
II. METHODS
The development of chaos theory in the last decades sup-
plied the framework to study nonlinear dynamics through a new
approach. Many efforts were made to explain the complex be-
havior of deterministic systems with the presence of nonlinear
120 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 1, JANUARY 2006
features instead of the usual stochastic models. The most direct
link between chaos theory and the real world is the analysis of
time series from real systems in terms of nonlinear dynamics [7].
In some situations, when the a priori knowledge about the gen-
erating system is deep enough, it is possible to precisely charac-
terize the structure of the system itself from the observed time
series. In these cases, the study involves a mathematical charac-
terization of the phase space or the estimation of some descrip-
tors of the attractor. This approach leads to the reconstruction of
the attractor in an embedding space, estimating some invariants
such as fractal dimension and Lyapunov exponents [8]–[11].
Nevertheless, it is often impossible to obtain sufficient in-
sight on biological systems from signals recorded by noninva-
sive techniques. The knowledge we have on their behavior is
limited by their intrinsic complexity, which is the result of nu-
merous interacting mechanisms contributing to the physiolog-
ical performance. The concept of complexity does not corre-
spond to a unique and rigorous definition. Sometimes it is re-
lated to the difficulties we have in describing or understanding
a signal.
Pincus [1] observed that the classic Kolmogorov-Sinai en-
tropy assumes infinite value for processes with superimposed
stochastic noise. Thus, it is unable to distinguish processes that
differ in complexity. In this case, the word complexity refers
to the unpredictability of the system state location, once the ini-
tial conditions are known. The less predictable the states are, the
more complex the system is. The association of unpredictability
with complexity leads to the conclusion that Gaussian noise is
very complex because all points in the phase space are a prob-
able state of the system and the state location is unpredictable.
In contrast, a periodic dynamical system with period 2 owns an
extremely low complexity level: in fact, all trajectories meet two
points with probability 1. Nevertheless most biological systems
lie between the two extremes of this complexity scale.
Starting from these considerations, Pincus proposed a family
of statistics, called Approximate Entropy [2], aimed at mea-
suring the signal regularity i.e., the presence of similar patterns
in the time series.
Given N data points ; 1 , the algorithm
constructs sequences and it computes, for each
1, the quantity
number of
(1)
measures, with a tolerance , the regularity of patterns
by comparing them to a given pattern of length ( and are
fixed values: is the detail level at which the signal is analyzed,
is a threshold, which filters out irregularities).
The regularity parameter is defined as
ApEn
where
1 tion of the factor
ApEn is the estimator of this
parameter for an experimental time series of fixed length N.
As Pincus himself noticed [1], [2], [12], approximate entropy
is affected by a bias effect. In addition, ApEn strongly
depends on the record length. The correction of the bias effect
is not trivial since the straightforward removal of self-counting
could origin a high sensitivity to outliers. Furthermore, Richman
and Moorman [13] demonstrated that in some cases ApEn lacks
of consistency. Its computation in irregular time series is af-
fected by a bias, which may cause an overestimate of the en-
tropy value, mostly when values are very small.
However, the application of ApEn statistics to real data often
obtains good results, giving general information about the reg-
ularity and the persistence of the signal. This fact has led to an
extensive use of ApEn in clinical applications, e.g., the analysis
of inter-beat RR signals in the study of cardiovascular diseases
[14]–[16].
Entropy methods exploit a symbolic representation of the
time series. Despite the severe reduction of information, they are
able to enhance relevant features of the signal. ApEn, and more
generally coarse-grained entropies, can be useful to track quali-
tative changes in time series patterns, without precisely charac-
terizing the generating system.
Richman and Moorman [13], [17] developed a modification
of the ApEn algorithm, named sample entropy (SampEn), in
order to remove some of the deficiencies reported here.
The differences with respect to ApEn are: 1) self-matches are
not counted; 2) only the first N- vectors of length are con-
sidered; and 3) the conditional probabilities are not estimated in
a template manner. Thus, the probability measure is computed
directly as the logarithm of conditional probability instead of
the ratio of the logarithmic sums.
SampEn shows a relative consistency in cases where ApEn
does not. Furthermore the values of SampEn agree with the the-
oretical values expected for a uniform random noise time series
much more than the ApEn values, even for very short time se-
ries [13].
Both ApEn and SampEn supply a single index concerning
the general behavior of the time series, but they cannot reveal
the underlying dynamics of the generating system. Thus, if the
signal X has a lower entropy value than the signal Y, we can
only say that X is more regular than Y. If the original purpose
of entropy was to identify chaotic dynamics, the statistics ApEn
and SampEn have changed the perspective by providing a figure
related to the regularity of the time series at the original time
scale.
Several methods have been proposed in the literature, for
achieving a thorough analysis at different scales. Among them
wavelet-transform modulus-maxima method (WTMM) appears
a very promising tool [18]. Other methods, calculating the Hurst
exponent estimate long range dependence scaling exponent and
statistical self-similarity as well as the presence of long range
correlation and “memory” in the process.
Costa et al. [3], [4] introduced the so-called MSE method.
The procedure considers a time series of N points and con-
structs consecutive coarse-grained time series , as a func-
1
1 (2)
FERRARIO et al.: COMPARISON OF ENTROPY-BASED REGULARITY ESTIMATORS 121

is the original time series, is the length of each

coarse-grained time series.
An entropy measure is calculated for each sequence
and it is plotted as function of the scale factor . The rationale
beyond this procedure is an enhancement of time series repeti-
tive patterns as a function of different scales. Differences in the
index at different time scales could help understanding the time
series in terms of regularity and structure (i.e., short versus long
range).

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 ).

time series 1000 . The persistence is coherent with the na-

ture of 1/f noise, showing strong correlations on long term. Fur-
thermore we obtained similar MSE values by using both ApEn
and SampEn estimators.
Fig. 2 reports the results of MSE analysis performed on
signals generated by completely different dynamical systems
(white noise and logistic map), confirming that the multiscale
analysis is strongly related to different signal structures. For
instance random noise has the highest value at scale factor 1
(original time series), but it shows a decreasing trend for other
scale factors. On the contrary, the time series generated by the
logistic map has an increasing trend presenting a maximum
at scale factor 4. Furthermore, as the figure shows, the ApEn
and SampEn values are very similar. Fig. 3 reports the results
on time series obtained from MIX process (the definition is
reported in the figure caption). The analysis shows that the
multiscale approach evidences periodic behaviors in the signal.
In fact, Fig. 3 evidences an abrupt decrease of entropy occurring
at scale factor 6 and 8, where the scale factor corresponds to a
submultiple of the signal period.
IV. APPLICATION TO THE FETAL HEART RATE SIGNAL
Our database includes more than 800 FHR traces, which
were collected from antepartum CTG recordings in 572 preg-
nant women. The data for each subject consist of the FHR trace,
complete anamnestic information and antepartum prognostic
indications. Data collection involved university departments
and a private company in Italy, in a joint research effort.
The women were in a gestational age ranging from 26th and
40th week. Recordings came from four identical devices (HP
M1350A), located in various university clinics in Rome. Each
recording lasted at least 60 min and it contains both the FHR
series and the toco trace. The HP M1350A was set to provide
two FHR samples/s, resulting in sequences of more than 7200
points. For a subset of CTG recordings, belonging to women
122 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 1, JANUARY 2006
6
Fig. 3. Results of the MSE analysis (avg. std.) for simulated signals, number
0 1 1
of points N = 7000. The MIX(p,N) is defined as (1 z ) x + z y , where z is a
random variable, that assumes value 1 with probability p and 0 with probability
p0
1 p p
p, x is a sequence of period T = 12 generated by the equation x =
2 sin(2j=12) and y is a uniformly distributed variable on [ 0 3; 3] The
lower p is chosen, the more periodic and regular the signal is.
who gave birth in those clinics, we were able to acquire outcome
data at delivery (weight of the newborn, type of delivery, Apgar
score, and final diagnosis).
In this paper, we considered two groups of fetuses: a control
group of Normal fetuses (absence of pathologies at birth, no an-
tepartum problems and spontaneous delivery) and a group of
Suffering fetuses, presenting different conditions of FD at birth,
often associated with pathologies. Within the Suffering group
we made a further subdivision by separating the recognized suf-
fering fetuses (RecSuf) from the not recognized suffering fetuses
(NotRecSuf). The former group showed some risk indications in
the antenatal period, at the moment of the CTG recording; the
latter contained fetuses that were suffering during labor or at
delivery, but were judged as absolutely normal at the moment
of the CTG recording. This particular subdivision allows sepa-
rating chronic distressed fetuses, which are commonly identified
in the antepartum period, from fetuses affected by acute or sub
acute suffering conditions arising during labor.
Each group (RecSuff, NotRecSuff and Normal) was further
subdivided by gestational age as shown in Table I.
The standard analysis procedure was carried out through the
identification of the baseline [20], the detection of accelerations
and decelerations, the computation of short-term variability and
long-term irregularity (LTI). A further power spectral density
(PSD) analysis was performed through software procedures ex-
pressly developed within our research project [21], [22].
During CTG monitoring, the fetus can move and breath
as well as contractions can occur. These events dramatically
change the FHR variability signal, demonstrating its nonsta-
tionary nature. Therefore, we calculated the parameters over
short sequences as suggested in a recent study [22]. This would
allow to capture the development of signal modifications and,
eventually, to understand physiological implications. In this
paper, we computed both the ApEn(1,0.1) and SampEn(1,0.1)
over sequences of 360 samples, corresponding to 3 min
TABLE I
CTG RECORDINGS SELECTED FOR THE ANALYSIS
All signals present a high quality of recording. Complete information both in
the antepartum period and at birth was available. (g.w.=gestational week). The
subset selected for the MSE analysis is reported in the last row.
recording. Then we considered, for each recording, the mean
value of entropy sequences, both for ApEn and SampEn mea-
sures. The resulting ApEn and SampEn indexes were compared
for the selected FHR signals referring to the defined groups.
The application of the multiscale method considered both
ApEn and SampEn with parameters 2, 0.15 . The
analysis was limited to scale factor 8 as the length of FHR
time series could be only 7200 point (1 h).
The MSE approach required a further selection of the avail-
able recordings by introducing a threshold on the FHR signal
quality. As a matter of fact, sometimes the FHR signal is not (or
very badly) detected by the ultrasound Doppler probe, and the
HP monitor attributes a zero value to signal loss. Thus, we im-
posed that FHR signal should contain less than 250 zeros. The
reason is that the MSE was computed on the whole FHR signal;
then, chunks of successive zeros might be interpreted as repeti-
tive patterns leading to inconsistent results. The introduction of
the quality threshold reduced the number of recordings, so we
did not maintain the subdivision on the basis of gestational age.
Details are reported in Table I.
Almost all time domain and morphological parameters are
not significantly different among the different classes. Only the
LTI index satisfies the imposed threshold 0.05 for the
Student t-Test, the Rank Sum Test and the Kolmogorov Smirnov
tests in the last class of gestational age ( 39 weeks) as reported
in Table II.
The PSD analysis does not show any significant difference
in the three considered frequency bands. However, we can ob-
serve that ApEn and SampEn are statistically significant
0.01 for the 30th–35th gestational weeks period (ANOVA test
and Kruskall–Wallis test performed within the three groups;
post-hoc comparisons by Scheffè test) and both provide similar
results, confirming their potential as a discriminating parameter.
Both ApEn and SampEn are higher in suffering fetuses than in
normal ones (see detials in Table II).
Furthermore the early class of suffering fetuses presents a
high percentage of preterm labor (most of them did not reach the
35th week). So the obtained result would confirm that, during
this gestational period, the episodes of serious distress often in-
duce dangerous situations for the fetus, since it is not yet com-
pletely developed and its regulation system fails in responding
to severe hypoxia.
FERRARIO et al.: COMPARISON OF ENTROPY-BASED REGULARITY ESTIMATORS 123

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.

Fig. 4. Results of multiscale analysis (avg. 6 std.). The squares represent

the normal fetuses (65 recordings) while the triangles refer to the suffering
fetuses, which have been classified as pathologic antepartum and were classified
as normal at delivery (8 recordings).
In the MSE analysis as reported here, we did not maintain the
subdivision in gestational ages because of the reduced number
of valid recordings. Globally MSE analysis provides similar re-
sults for both entropy estimators (ApEn, SampEn). We observe
no significant differences between normal fetuses and suffering
fetuses without pathology at birth, as shown in Fig. 4 (i.e., the
fetuses that were judged as suffering in the antepartum period
but they resulted healthy at birth). On the other hand, significant
differences are found between normal and suffering fetuses pre-
senting a pathology at birth values 0.01 . In this case, all
MSE indexes are higher for normal fetuses as shown in Fig. 5.
V. DISCUSSION AND CONCLUSION
The multiscale entropy analyses of simulated signal provide
interesting indications for the study of time series. While MSE is
an overall measure of regularity, the multiscale approach gives
more details about the time scales at which irregularities occur.
It allows extracting the information about the signal structure;
for example if the irregularity belongs to the first scale factor
only (e.g., white noise) or it persists for all scale factors (e.g.,
1/f noise).
As it normally happens with entropy based estimators, it is
sometimes difficult to link the index behavior to a specific phys-
ical system component. This is certainly a limit of these methods
Fig. 5. Results of multiscale analysis (avg. 6 std.) The squares represent
the normal fetuses (65 recordings) while the triangles represent the suffering
fetuses, which have been classified antepartum as pathologic (30 recordings).
The diagnosis was confirmed at delivery.
that is rewarded by their robustness against noise and by the ab-
sence of a priori hypothesis on the time series structure [20].
This is the case of other well-known estimators as the wavelet-
based scaling exponent indices. [18]. Moreover, it is quite evi-
dent that the index value depends on the choice of the symbolic
encoding (i.e., the choice of the coarse graining and of the pa-
rameters characterizing the entropy index). An analogy can be
found when we apply short term (3 min, 360 points) and long
term spectral analysis to HR variability. Short-term analysis ev-
idences harmonic contributions involved in the short time regu-
lation. Mechanisms controlling the heart in the long period have
a completely different pattern (1/f like behavior).
A. Discussion of FHR experimental results
The standard time domain and morphological parameters
confirm their weakness in detecting or identifying fetal danger,
as already demonstrated in many studies [23]. In this paper,
these techniques are not able to distinguish between normal
fetuses and distressed ones. The LTI parameter only seems
to be significantly different for the gestational class 39
weeks, but most fetuses presenting severe distress problems
do not reach this stage of gestation because normally, in these
cases, clinicians induce a premature delivery. On the contrary,
regularity analyses both in short (single scale) and in long
124 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 1, JANUARY 2006
(multiscale) time windows are able to significantly discriminate
the true distressed and pathological fetuses from the healthy
and not suffering ones, although MSE analysis and single
ApEn/SampEn index apparently seem to provide contradictory
results.
The MSE values of normal fetuses increase, as the time se-
ries is progressively coarse grained. In this case, values ob-
tained in distressed fetuses show that their HRV signal has a
different signal structure (more regular, with a nearly sinusoidal
behavior) than normal FHR, which reveals an increasing trend
that is distinctive of more irregular signal patterns. Comparison
with results of MIX(p) processes reinforces this hypothesis.
The MSE approach for the whole recording time (7200
points) shows that pathological fetuses present lower entropy
values than healthy ones at all scale factors, supporting the idea
that, as in adults, pathological conditions induce an increase of
signal regularity in the long-term, due to a possible loss of com-
plexity (information reduction) in the regulatory mechanisms.
The results of the MSE approach show that FHR variability
possesses a scale organization, which can be only highlighted
by considering the entire signal. In fact, we observe that the
MSE SampEn at scale 1 (considering all the 7200 points) is
lower than we previously obtained over 360 points.
The results of both entropies (ApEn and SampEn), computed
as a single averaged index over short time windows (360 points),
show an opposite behavior with respect to the multiscale ap-
proach. Their interpretation is not trivial, and can be found by
considering both the mathematical characteristics of entropy es-
timators over short time sequences and the development of FD
during pregnancy.
ApEn and SampEn resulting from short FHR time series in
the original time scale are more sensitive to the local irregular-
ities (local differences) of signal patterns. In fact, as it could be
expected, both ApEn and SampEn show higher values than in
the multiscale approach. As a matter of fact, the multiscale anal-
ysis over the entire signal does not enhance the abrupt events,
which are filtered out by the general trend of the signal for all
scale factors (even for factor 1, thus, the same scale used in short
time series). The ApEn or SampEn values computed over short
windows provide instead the “local” information content corre-
lated to the events.
The inversion of entropy levels in a short time analysis should
not be surprising. Sometimes it happens that both ApEn and
SampEn algorithms, when applied to some pathologic time se-
ries, provide higher entropy values than those obtained from free
running physiologic systems under healthy condition [19].
In relation to our FD problem, a possible interpretation of this
experimental evidence can be found by comparing the obtained
results with those reported in a previous work [24] related to
normal fetuses submitted to vibroacoustic stimulation. In that
work, when the fetus was in a quiet period before the stimulus
was given, the ApEn values were higher than in the activity pe-
riod after the vibroacoustic stimulation. We can, thus, make the
reasonable hypothesis that higher entropy values found in suf-
fering fetuses by averaging ApEn values, computed on succes-
sive 3-min windows for the whole 1-h recording, can be due to
a prevalence of quiet periods in suffering fetuses, with respect
to a prevalence of activity periods in healthy fetuses analyzed at
the same gestational age.
At this stage of gestation (30–35 weeks), in suffering fetuses,
the neural regulatory mechanisms do not yet show a clear al-
teration but the prevalence of quiet periods in the FHR could
be a marker of increasing distress conditions which could lead,
during the ongoing pregnancy, to pathology of the fetus.
Longer quiet periods might mean that the disease is starting
but it does not yet affect the neural regulatory mechanisms
controlling the heart frequency. The consequence is an initial
loss of awakeness of the fetus (higher entropy) before a clear
pathology emerges, changing the behavior of regulatory mech-
anisms (lower entropy, as it happens in the class of fetuses with
pathology at birth).
In conclusion, entropy indexes show their ability in inves-
tigating the structure of biological time series. FHR analysis
points out that the entropy estimators are sensitive to short-term
and long-term signal patterns. In the short term, they catch local
irregularities as the dominant mechanism. The long term anal-
ysis draws advantage from the severe reduction of information
which is imposed by the scaling procedure to enhance relevant
features in the FHR signal. Results evidence a quite marked de-
tection of severely distressed fetuses on the basis of differences
in the variability signal structure.
These considerations reinforce the interest toward the use of
advanced and nonconventional signal processing methods in the
clinical routine. The discriminating ability of regularity mea-
sures enhances the emergence of information with biological
importance from the time series structure.
ACKNOWLEDGMENT
The authors would like to thank Prof. D. Arduini for his sup-
port in the data collection process at the Ob-Gyn Clinic, Hospital
“Fatebenefratelli,” Rome, Italy, and Dr. S. Ramat for revising
our paper.
REFERENCES
[1] S. M. Pincus, “Approximate entropy (ApEn) as complexity measure,”
Chaos, vol. 5, no. 1, pp. 110–117, Mar. 1995.
[2] S. M. Pincus, I. M. Gladstone, and R. A. Ehrenkranz, “A regularity
statistic for medical data analysis,” J. Clin. Monit., vol. 7, no. Oct., pp.
335–345, 1991.
[3] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy to
distinguish physiologic and synthetic RR time series,” Comput. Cardiol.,
vol. 29, pp. 137–140, 2002.
[4] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy anal-
ysis of complex physiologic time series,” Phys. Rev. Lett., vol. 89, no. 6,
p. 068 102, Aug. 2002.
[5] H. M. McNamara and G. A. Dildy, “Continuous intrapartum pH, pCO2,
SpO2 monitoring,” Obstet. Gynecol. Clin. N. Am., vol. 26, pp. 671–693,
Dec. 1999.
[6] P. Malcus, “Antenatal fetal surveillance,” Curr. Opin. Obstet. Gynaecol.,
vol. 16, pp. 123–128, 2004.
[7] H. Kantz and T. Schreiber, Nonlinear Time Series Analysis, 2 ed. Cam-
bridge, U.K.: Cambridge Univ. Press, 2004.
[8] T. S. Parker and L. O. Chua, “Chaos: a tutorial for engineers,” Proc.
IEEE, vol. 75, no. 8, pp. 982–1008, Aug. 1987.
[9] E. Ott, Chaos in Dynamical Systems. New York: Cambridge Univ.
Press, 1993, pp. 1–22.
[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.
[11] P. Grassberger and I. Procaccia, “Estimation of the Kolmogorov entropy
from a chaotic signal,” Phys. Rev. A, vol. 28, pp. 2591–2593, May 1983.
FERRARIO et al.: COMPARISON OF ENTROPY-BASED REGULARITY ESTIMATORS
[12] S. M. Pincus, “Approximate entropy as a measure of system complexity,”
in Proc. Nat. Acad. Sci. USA, vol. 88, Mar. 1991, pp. 2297–2301.
[13] J. S. Richman and J. R. Moorman, “Physiological time-series analysis
using approximate entropy and sample entropy,” Am. J. Physiol. Heart
Circ. Physiol., vol. 278, pp. H2039–2049, Jun. 2000.
[14] T. H. Mäkikallio et al., “Dynamic analysis of heart rate may predict
subsequent ventricular tachycardia after myocardial infarction,” Am. J.
Cardiol., vol. 80, no. 6, pp. 779–783, Sep. 1997.
[15] K. K. L. Ho et al., “Predicting survival in heart failure case and control
subjects by use of fully automated methods for deriving nonlinear and
conventional indexes of heart rate dynamics,” Circulation, vol. 96, no.
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
of neonatal heart rate variability,” Am. J. Physiol. Regul. Integr. Comp.
Physiol., vol. 283, pp. R789–R797, Sep. 2002.
[18] B. Audry, E. Bacry, J.-F. Muzy, and A. Arneodo, “Wavelet-based esti-
mators of scaling behavior,” IEEE Trans. Inform. Theory, vol. 48, no.
11, pp. 2938–2954, Nov. 2002.
[19] M. Costa, A. L. Goldberger, and C.-K. Peng, “Multiscale entropy anal-
ysis of biological signals,” Phys. Rev. E, vol. 71, p. 021 906, 2005.
[20] R. Mantel, H. P. van Geijn, F. J. Caron, J. M. Swartjes, E. E. van Wo-
erden, and H. W. Jongsma, “Computer analysis of antepartum fetal heart
rate: 1. Baseline determination,” Int. J. Biomed. Comput., vol. 25, no. 4,
pp. 261–272, 1990.
[21] G. Magenes, M. G. Signorini., D. Arduini, F. Lunghi, and A. Bonalumi.
Development of a new system for cardiotocographic automatic analysis.
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.
[23] H. P.Van-Geijn , “Developments in CTG analysis,” Baillieres Clin. Ob-
stet. Gynaecol., vol. 10, no. 2, pp. 185–209, 1996.
[24] G. Magenes, M. G. Signorini, D. Arduini, and S. Cerutti, “Fetal Heart
rate variability due to vibroacoustic stimulation: linear and nonlinear
contribution,” Meth. Inf. Med., vol. 43, no. 1, pp. 47–51, 2004.
Manuela Ferrario received the Italian degree
(Laurea) from the Polytechnic University, Milano,
Italy, in 2003. She received a scholarship for the
development of a prototype clinical decision support
system and the improvement of the current decision
scores for a correct and early evaluation of fetal
distress. She is currently working towards the Ph.D.
degree in the Biomedical Engineering Department
at the same university.
Her main research interests are nonlinear analysis,
fetal heart rate analysis, and multiscale and fractal
analysis.
125
Maria G. Signorini received the Ph.D. degree in
1995 from the Polytechnic University, Milano, Italy.
She is currently an Associate Professor in the
Biomedical Engineering Department at the same
University. Her main research interests are nonlinear
analysis and modeling of biological signals and
systems regarding, in particular, the cardiovascular
system dynamics. She cooperates in industrial
projects (HP Italy, Sister-Fresenius Medical Care)
on biomedical signal processing and instrumentation
improvement.
Giovanni Magenes received the “Laurea” degree
in electronic engineering from the University of
Pavia, Pavia, Italy, in 1981 and the Ph.D. degree
in bioengineering from the Polytechnic of Milano,
Milano, Italy, in 1986.
Since 1988, he is with the Dipartimento di
Informatica e Sistemistica, University of Pavia,
where he is now full Professor of Biomedical Signal
Processing. His main interests are in intelligent
biosignal analysis, neural networks applications in
robotics and biomedicine, and sensorimotor control
in men and machines.
Dr. Magenes is member of the European Neural Networks Society, and the
Barany Society.
Sergio Cerutti (M’81–SM’97–F’02) is Professor
of Biomedical Signal and Data Processing at the
Department of Biomedical Engineering of the
Polytechnic University of Milan, Milan, Italy.
Since November 2000, he is the Chairman of the
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
potentials), cardiovascular modeling, neurosciences,
medical informatics, and regulation and standards
in medical equipments and devices. Since March 1983, he has also taught a
course at a graduate level on Biomedical Signal Processing at Engineering
Faculties (Milan and Rome) and at Specialization Schools of Medical Faculties
(Milan and Roma).
Prof. Cerutti was an elected member of IEEE-EMBS AdCom (Region 8)
for the period 1993–1996. He is an Associate Editor of IEEE TRANSACTIONS
BIOMEDICAL ENGINEERING. He is a member of the Steering Committee of
the IEEE-EMBS Summer School on Biomedical Signal Processing; he was
the local organizer of three IEEE-EMBS Summer Schools held in Siena in
1995/1999/2003.