Dispersion Entropy: A Measure of Electrohysterographic Complexity
for Preterm Labor Discrimination
Félix Nieto-del-Amor
1a
, Yiyao Ye-Lin
1b
,
Javier Garcia-Casado
1c
, A. Diaz-Martinez
1d
,
María González Martínez
2
, R. Monfort-Ortiz
2
and Gema Prats-Boluda
1e
1
Centro de Investigación e Innovación en Bioingeniería, Universitat Politècnica de València, 46022 Valencia, Spain
2
Servicio de Obstetricia, H.U. P. La Fe, 46026 Valencia, Spain
Keywords: Dispersion Entropy, Sample Entropy, Preterm Birth, Electrohysterography, EHG.
Abstract: Although preterm labor is a major cause of neonatal death and often leaves health sequels in the survivors,
there are no accurate and reliable clinical tools for preterm labor prediction. The Electrohysterogram (EHG)
has arisen as a promising alternative that provides relevant information on uterine activity that could be useful
in predicting preterm labor. In this work, we optimized and assessed the performance of the Dispersion
Entropy (DispEn) metric and compared it to conventional Sample Entropy (SampEn) in EHG recordings to
discriminate term from preterm deliveries. For this, we used the two public databases TPEHG and TPEHGT
DS of EHG recordings collected from women during regular checkups. The 10
th
, 50
th
and 90
th
percentiles of
entropy metrics were computed on whole (WBW) and fast wave high (FWH) EHG bandwidths, sweeping the
DispEn and SampEn internal parameters to optimize term/preterm discrimination. The results revealed that
for both the FWH and WBW bandwidths the best separability was reached when computing the 10
th
percentile,
achieving a p-value (0.00007) for DispEn in FWH, c = 7 and m = 2, associated with lower complexity preterm
deliveries, indicating that DispEn is a promising parameter for preterm labor prediction.
1 INTRODUCTION
A preterm birth (PB) is a high-risk situation and has a
prevalence of up to 10% of all labor cases, affecting
more than 15 million families worldwide (Fuchs et
al., 2004). The consequences of PB affect maternal-
fetal health and is the main cause of mortality in
children under 5 years of age (Leung, 2004). It also
has a high economic cost for national healthcare
systems (Petrou et al., 2019). There are now several
techniques available for preterm birth detection,
mainly the measure of cervical length (O’Hara et al.,
2013) and biochemical markers (Leung, 2004).
However, while these techniques provide a highly
negative predictive value, their positive predictive
values are quite low and do not identify preterm
deliveries (Berghella et al., 2008; Diaz-Martinez et
al., 2020).
a
https://orcid.org/0000-0003-0050-9360
b
https://orcid.org/0000-0003-2929-181X
c
https://orcid.org/0000-0003-1410-2721
d
https://orcid.org/0000-0002-4605-6048
e
https://orcid.org/0000-0002-9362-5055
Intrauterine pressure catheter IUPC and
tocodynamometry TOCO are the classical methods of
measuring uterine dynamics. However, the former is
an invasive method and involves risks, while the
latter, although non-invasive, is neither very sensitive
or precise (Euliano et al., 2016). The aim of the
electrohysterographic technique is to deal with these
limitations. The electrohysterogram (EHG) is the
bioelectric signal recording of the muscular activity
of the myometrium. The generation and propagation
of action potentials through a suitable number of
myometrial cells induce uterine muscle contractions
and raise the internal uterine pressure. EHG can
provide essential information on uterine activity
(Devedeux et al., 1993). EHG energy is distributed
between 0.1 and 4Hz and is composed of two waves.
The slow wave (SW) has a period equal to the
duration of contraction and as its bandwidth overlaps
260
Nieto-del-Amor, F., Ye-Lin, Y., Garcia-Casado, J., Diaz-Martinez, A., Martínez, M., Monfort-Ortiz, R. and Prats-Boluda, G.
Dispersion Entropy: A Measure of Electrohysterographic Complexity for Preterm Labor Discrimination.
DOI: 10.5220/0010316602600267
In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 4: BIOSIGNALS, pages 260-267
ISBN: 978-989-758-490-9
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
the baseline wander it is difficult to analyze and
extract reliable information from it. The fast wave
(FW) is superimposed on the slow wave signal and
can be divided into two parts according to the
frequency in which it is presented: fast wave low
(FWL), whose frequency peak is between 0.13 and
0.26 Hz and is supposed to be associated with
contraction propagation and fast wave high (FWH),
whose frequency peak is between 0.26 and 0.88 Hz
and is related to uterine cell excitability (Devedeux et
al., 1993; Terrien et al., 2007)
Previous studies have shown that EHG signals
include relevant information on the state of the
uterine electrophysiology and are better able to detect
real preterm cases (Devedeux et al., 1993). It has been
shown that as labor approaches synchronization,
amplitude and predictability increase, complexity is
reduced and there is a shift of frequency content to
high frequencies (Devedeux et al., 1993; Garfield &
Maner, 2007). The literature proposes different non-
linear parameters, such as entropy metrics, to
characterize EHG. These latter reduce their values
when regularity increases (Mischi et al., 2018).
Sample entropy (SampEn) (Richman and
Moorman, 2000) is a widely used metric to
discriminate preterm from term (Garcia-Casado et al.,
2018). Studies in the literature suggest that SampEn
performs best in discriminating term vs preterm
deliveries in EHG recordings among the non-linear
parameters tested (maximal Lyapunov exponent and
correlation dimension) and even outperformed
spectral parameters such as median frequency (Fele-
Žorž et al., 2008; Rostaghi and Azami, 2016). In
addition, SampEn have been used to analyze the
feasibility of using EHG to discriminate women with
threatened preterm labor who gave birth in less than
7 days from those who delivered in more than 7 days
(Mas-Cabo, Prats-Boluda, Perales, et al., 2019).
Rostaghi & Azami (2016) proposed Dispersion
entropy (DispEn) to deal with the limitations of
SampEn, which is sensitive to changes in
simultaneous frequency and amplitude. DispEn is a
computationally efficient measure of the regularity of
time series and outperforms entropy metrics for
characterizing other biomedical signals (Rostaghi and
Azami, 2016). The performance of DispEn has been
explored in different biomedical signals, but in EHG
not yet. Azami et al analyzed magnetoencephalogram
MEG signals to discriminate Alzheimer’s disease
patients from an elderly control group. When
SampEn, permutation entropy (PermEn) (Bandt &
Pompe, 2002), fuzzy entropy (de Luca & Termini,
1993) and DispEn were computed for MEG signals,
DispEn presented the highest separability between
the groups than other entropy metrics(Azami,
Rostaghi, et al., 2016). Kafantaris et al used DispEn
to characterize electrocardiogram ECG signals with
different types of heartbeat (normal-healthy hearbeats,
atrial premature beats and premature ventricular
contractions) and concluded that the algorithm is
capable of producing significantly different DispEn
distributions between the different groups of
heartbeats (Kafantaris et al., 2019). Rostaghi &
Azami studied electroencephalographic signals with
DispEn to test their ability for discriminating focal
and non-focal EEG signals and found that DispEn had
better separability than SampEn and PermEn
(Rostaghi and Azami, 2016). All these case studies
suggest that DispEn is a good estimator of signal
regularity and improves the separability of the groups
under study.
In the present work, we attempted to optimize,
assess and compare DispEn performance with
SampEn to discriminate EHG recordings between
term and preterm deliveries during routine checkups,
when computed in the fast wave high and in the whole
EHG bandwidth.
2 MATERIALS AND METHODS
2.1 EHG Data Bases
Two public EHG data bases available in Physionet
were used for the case study, the “Term-Preterm EHG
Database” (TPEHG) (Fele-Žorž et al., 2008) and the
“The Term-Preterm EHG Dataset with tocogram”
(TPEHGT DS) (Jager et al., 2018) Both have been
widely used in comparative studies on term and
preterm cases and were obtained by the Department
of Obstetrics and Gynecology at the Ljubljana
University Medical Center.
A total of 326 EHG signals with 275 term labor
cases (labor > 37 weeks) and 51 preterm labor cases
(labor < 37 weeks) were recorded during routine
checkups of pregnant women between 22 and 37
weeks of gestation. No induced labor cases or
caesarean deliveries were included.
Thirty-minute EHG signal were recorded in both
databases with the same protocol, consisting of four
disposable electrodes on the woman’s abdomen at an
interelectrode distance of 7cm (Fele-Žorž et al., 2008).
Three bipolar channels, S1, S2 and S3, were obtained
after removing the monopolar EHG recordings, as
shown in Figure 1. Previous studies pointed out that
EHG features from S3 outperform S1 and S2
channels in distinguishing term and preterm
deliveries (Fele-Žorž et al., 2008; Mas-Cabo, Prats-
Dispersion Entropy: A Measure of Electrohysterographic Complexity for Preterm Labor Discrimination
261
Boluda, Garcia-Casado, et al., 2019). Therefore, only
channel S3 was analyzed in the present study. The
bipolar signals were digitized at 20 samples per
second, with a resolution of 16 bits and over a range
of ±2.5 millivolts (Fele-Žorž et al., 2008).
2.2 EHG Signal Characterization
The EHG recordings were filtered in the range 0.1 to
4 Hz by a zero-phase shift 5
th
order Butterworth
bandpass, since this bandwidth comprises the main
content of EHG signals.
Figure 1: Recording protocol of EHG signals. Modified
from (Jager et al., 2018).
Segments with motion artifacts were visually
identified by experts and discarded from the study.
The criteria adopted to discard EHG sections were:
non-physiological events with a significant abrupt
increase in amplitude compared to basal activity, and
respiratory interference associated with the
appearance of periodic components with frequencies
in the band of 12 and 24cpm (0.2–0.4Hz). 220 term
and 40 preterm EHG recordings were analyzed.
SampEn and DispEn parameters were computed in
120s EHG signal analysis windows with a 50%
overlap (Mas-Cabo, Prats-Boluda, Perales, et al.,
2019) to keep the EHG section at a reasonable
computational cost with the minimal loss of
information (Azami & Escudero, 2018). This analysis
does not require contraction segmentation, which can
be tedious and subjective, and is more suitable for
future clinical use in real-time applications. SampEn
and DispEn were computed for both whole EHG
bandwidth (0.1-4Hz, WBW) and in the Fast Wave
High bandwidth (0.34-4Hz, FWH).
2.2.1 Sample Entropy
SampEn is the negative natural logarithm of the
probability that two similar sequences for m points in
the time series remain similar at the next point, in
which self-matches are not included in calculating the
probability (Richman and Moorman, 2000), so that a
lower SampEn value also indicates more self-
similarity in the time series. SampEn is an
improvement of approximate entropy (Pincus, 1991)
and is a frequently used metric in EHG to distinguish
between term and preterm cases for preterm birth
detection (Fele-Žorž et al., 2008).
SampEn depends on two internal parameters: the
length m of the templates to be compared constructed
from the time series, and a filtering threshold r, the
tolerance of mismatch between the corresponding
elements of the templates. Typically, the value of r is
considered as 0.15 to 0.25 times the value of standard
deviation (SD) of the time series, avoiding most of the
noise present in it. The value of m may be taken
considering that the length of the time series is
between 10
m
and 10
m+1
, although this latter is not so
restrictive (Xiong et al., 2019).
For the present work r was considered as 0.2
times the value of SD and sweeping m from 2 to 5.
2.2.2 Dispersion Entropy
Rostaghi & Azami (2016) proposed a new irregularity
indicator termed dispersion entropy (DispEn) based
on symbolic dynamics or patterns, transforming data
into a new signal with only a few different patterns
and simplifying the study of dynamic time series to a
distribution of symbol sequences. It was aimed at
dealing with the shortcomings of other entropy
parameters such as SampEn and PermEn. Thus,
unlike other entropy metrics, DispEn is sensitive to
changes in simultaneous frequency and amplitude
values and discriminate diverse biomedical and
mechanical states (Azami and Escudero, 2018).
DispEn depends on the mapping function and
three internal parameters: the length m of templates;
the number of classes c that determine the number of
patterns or classes to be considered in the
computation, and time delay d. It is recommended to
assume d = 1 for the latter parameter and for m and c
consider c
m
< N, N being the length of the time series.
If c is too low, always with c > 1, signal values are
too far and it leads to being assigned to the same class,
while if c is too high, small variations in the signal
can cause a change of class, making it sensitive to
noise (Rostaghi & Azami, 2016). Taking these
considerations into account, in the present work c was
swept between 3 and 9, m between 2 and 5 and d fixed
to 1. The mapping functions considered were: linear
BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing
262
mapping (linear), normal cumulative distribution
function (NCDF), tangent sigmoid (tansig),
logarithm sigmoid (logsig) and sorting method (sort).
2.2.3 Feature Extraction
A previous study revealed that the 10
th
and 90
th
percentiles outperform the 50
th
percentile of EHG
parameters for differentiating between different
obstetric scenarios, including preterm versus term
delivery (Mas-Cabo et al., 2020). We thus aimed to
consider the contractile activity present in the 120s
window analysis without segmenting the EHG
recordings. For complexity parameters such as
SampEn, the 10
th
percentile has been found to
perform best in discriminating term and preterm labor
from EHG registers. We therefore worked out the 10
th
,
50
th
and 90
th
percentiles of SampEn and DispEn
distributions.
2.2.4 Statistical Analysis
The Wilcoxon Rank-Sum Test was performed to
compare SampEn and DispEn’s ability to distinguish
term and preterm deliveries from EHG recordings in
routine checkups. This is a non-parametric statistical
hypothesis test used to compare two related samples
to assess whether their population mean ranks
differ( < 0.05). It is also suitable for non-normal
distributions such as SampEn and DispEn (see
Figures 2-3).
3 RESULTS
Figures 2 and 3 show the box and whisker plots of the
of 10
th
, 50
th
and 90
th
percentiles of EHG SampEn and
DispEn metrics for term and preterm groups
considering the combination of internal parameter
sweeps mentioned in Materials and Methods. As in
the case of DispEN the different mapping functions
obtained similar results, only the values for sort
mapping are represented.
In Figure 2, it can be seen that preterm group
median values are lower than those of term for the
10
th
and 50
th
SampEn percentiles. This agrees with
other studies in the literature that found the shorter the
time to delivery the higher the predictability of the
EHG signal, and the lower the entropy value. In
addition, the median values seem to decrease as m
passes from 2 to 3.
Table 1 contains the p-values of the Wilcoxon
Rank-Sum Test of SampEn values for distinguishing
term and preterm groups for each entropy parameter
distribution considered. Only the 10
th
percentile of
SampEn showed statistically significant differences
(p-value < 0.05) between the term and preterm groups,
and the results from the 10
th
to 90
th
percentile got
worse as m increased. The p-values were lower in the
FWH than WBW.
Table 1: P-values (Wilcoxon Rank-Sum Test) for SampEn
computed with r = 0.2ꞏSD when comparing different term
and preterm groups. Significant differences (p-value <
0.05) are shaded and the minimum p-value for FWH and
WBW bandwidths are in bold.
m 2 3
10
th
FWH
0.00123 0.00964
WBW
0.01071 0.01432
50
th
FWH
0.08753 0.28320
WBW
0.20581 0.44725
90
th
FWH
0.83969 0.68999
WBW 0.82898 0.81121
Figure 2: Boxplot SampEn distributions from EHG signals in different bandwidths and percentiles.
Dispersion Entropy: A Measure of Electrohysterographic Complexity for Preterm Labor Discrimination
263
Figure 3: Box and whisker plot distributions of DispEn using sorting mapping function computed from EHG signals in
different bandwidths and percentiles.
The distribution of the 10
th
and 50
th
DispEn
percentiles had lower median values for the preterm
than term group, as shown in Figure 3, confirming, as
has been stated in the literature, that EHG signal
predictability increases as time to delivery decreases,
but this was not so clear for the 90
th
percentile. Also,
the DispEn median values decreased as the m and c
values increased. Although Figure 3 only shows the
distribution obtained for sort mapping function, it is
representative for the rest of the cases, because of
their similar trend presented.
Tables 2 and 3 shows the p-values of the
Wilcoxon Rank-Sum Test of DispEn values of the
sort and NCDF mapping function. For the sort
mapping (Table 2), statistically significant
differences between term and preterm groups were
only obtained for the 10
th
and 50
th
percentiles in FWH
bandwidth for both m = 2 and 3. However, the lowest
p-value was reached with m = 2 and c = 7 for the 10
th
percentile. There were thus no internal parameter
combinations for WBW that achieved statistically
significant differences (p-value < 0.05) between term
and preterm groups with this mapping function. In the
NCDF mapping function, p-values were lower than
0.05 in the 10
th
and 50
th
percentiles for FWH and in
the 10
th
percentile for WBW. If Table 2 and Table 3
are compared it can be seen that although significant
statistical results in WBW were reached using NCDF,
the p-values were lower for FWH using sort mapping.
The best discrimination between the groups was at m
= 2 and c = 7, in which DispEn reached its optimum
value in WBW and FWH, the latter having the lowest
p-value.
Comparing the SampEn and DispEn outcomes,
commonly is obtained that lower median values of the
preterm than the term group were obtained and the
median values of the 10
th
to 90
th
distributions
decreased as m increased. Both entropy measures
were best able to distinguish between term and
preterm groups (lowest p-values in the test of
Wilconxon) when m = 2 in the 10
th
percentile.
DispEn outperformed SampEn in discriminating
in FWH (p-value, 0.00007), suggesting a better
ability to separate term and preterm groups. However,
in WBW SampEn reached a lower p-value, 0.01071
than DispEn.
4 DISCUSSION
SampEn is considered as one of the most used non-
linear metrics to discriminate between term and
BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing
264
preterm cases. One of the facts which do this possible
is when SampEn is evaluated for women delivering
prematurely, it is obtained a lower median value than
those of term case (di Marco et al., 2014). This fact is
consistent with the outcomes reaches in this study for
SampEn, and in the same way, the mean values of
DispEn acquired similar distributions. Consequently,
DispEn might be considered as a replacement of
SampEn in the measure of complexity in EHG signals
for discriminate between term and preterm cases. Our
findings reveal that DispEn better distinguishes term
from preterm deliveries than SampEn when
computed on EHG signal sections obtained in routine
checkups, (p-value of 0.00007 vs 0.00123). The
optimization of internal DispEn parameters was
achieved for an embedding dimension of m = 2 and a
number of classes c = 7, for the FWH and 10
th
percentile. This result agrees with our previous work
that showed the 10
th
percentile of sample entropy and
Lempel-Ziv non-linear parameters also outperformed
the 50
th
for distinguishing term and preterm labor and
other obstetric scenarios (Mas-Cabo et al., 2020).
This suggests the possibility of characterizing
contractile activity present in 120s EHG analysis
windows without the need for segmentation.
However, other approaches should also be
considered. The present study was performed using
channel S3 only, which had outperformed channels
S1 and S2 in term/preterm discrimination in previous
studies (Ahmed and Mandic, 2011; Azami et al.,
2019; Azami, Smith, et al., 2016). In future work it is
aimed to extend this study by assessing the
Table 2: P-values (Wilcoxon Rank-Sum Test) for DispEn computed with sort mapping function when comparing different
term and preterm groups. Significant differences (p-value < 0.05) are shaded and the minimum p-value is in
bold.
Table 3: P-values (Wilcoxon Rank-Sum Test) for DispEn computed with mapping function NCDF when comparing different
term and preterm groups. Significant differences (p-value < 0.05) are shaded and the minimum p-value is in
bold.
m c 3 4 5 6 7 8 9
2
10
th
FWH 0.00015 0.00019 0.00012 0.00015 0.00018 0.00015 0.00016
WBW 0.02078 0.02040 0.02732 0.03194 0.01850 0.02797 0.03176
50
th
FWH 0.00550 0.00611 0.00414 0.00393 0.00438 0.00594 0.00558
WBW 0.27108 0.27912 0.30853 0.28115 0.28115 0.30314 0.31510
90
th
FWH 0.09229 0.13584 0.14194 0.12937 0.15148 0.15148 0.17418
WBW 0.70182 0.66823 0.70691 0.77772 0.80236 0.76897 0.76548
3
10
th
FWH 0.00020 0.00021 0.00017 0.00022 0.00022 0.00019 0.00020
WBW 0.02913 0.02732 0.03362 0.03381 0.02247 0.03051 0.03381
50
th
FWH 0.00914 0.00932 0.00655 0.00582 0.00620 0.00866 0.00920
WBW 0.23236 0.28525 0.29780 0.26515 0.27810 0.29044 0.30745
90
th
FWH 0.12538 0.17059 0.19301 0.20746 0.21751 0.22792 0.30962
WBW 0.72398 0.71713 0.72912 0.78826 0.77947 0.76199 0.74637
m c 3 4 5 6 7 8 9
2
10
th
FWH 0.00015 0.00014 0.00009 0.00008 0.00007 0.00009 0.00008
WBW 0.10883 0.08668 0.10053 0.09097 0.06042 0.07938 0.07516
50
th
FWH 0.00777 0.00692 0.00793 0.00849 0.00726 0.00772 0.00692
WBW 0.41908 0.39326 0.46521 0.41777 0.39579 0.39579 0.39199
90
th
FWH 0.07667 0.10783 0.11241 0.09679 0.10486 0.11189 0.11241
WBW 0.76199 0.80060 0.82720 0.85760 0.84684 0.84148 0.86299
3
10
th
FWH 0.00019 0.00016 0.00011 0.00015 0.00012 0.00013 0.00011
WBW 0.11138 0.09588 0.10148 0.08216 0.05856 0.07078 0.06168
50
th
FWH 0.01335 0.00938 0.01158 0.01121 0.00908 0.01121 0.01196
WBW 0.41384 0.40734 0.46381 0.40476 0.39707 0.37330 0.36116
90
th
FWH 0.09497 0.15811 0.16846 0.18607 0.17274 0.20094 0.21077
WBW 0.77597 0.77947 0.80944 0.85939 0.80060 0.84505 0.80944
Dispersion Entropy: A Measure of Electrohysterographic Complexity for Preterm Labor Discrimination
265
performance of multivariate DispEn and other
multivariate entropy algorithms (Mas-Cabo et al.,
2020) in multichannel EHG databases so as to define
an optimal set of features to develop a preterm labor
predictor.
Other point to be considered is that SampEn has
certain limitations, such as its higher computational
cost than other entropies in similar assessment
conditions and its undefined or unreliable results for
short signals (Azami & Escudero, 2018). For a real-
time or a fast post-processed application in preterm
birth prediction with EHG signals may be useful
provide of an algorithm capable to compute a
measure of complexity of the signal as faster as it is
possible. Thus, DispEn outperform SampEn and
other relative entropy metric in which computational
cost is referred (Azami and Escudero, 2018).
In addition, only a statistical approach of the
separability of probability distributions of term and
preterm registers is taken into account. However, for
summiting a robust preterm labor discriminator, not
only one metric is used. In this way, the
complementation with other metrics related to EHG
signals should be evaluated and so obtain if the
append of DispEn outperforms the scores obtained
with SampEn in similar conditions.
Although this work focused on term vs preterm
discrimination with EHG entropy metrics, EHG can
also be used in other obstetric scenarios (Mas-Cabo et
al., 2020). DispEn may be suitable for the prediction
of labor induction success (Benalcazar-Parra et al.,
2019) or detecting imminent delivery in women with
threatened preterm labor under tocolytic treatment
(Mas-Cabo, Prats-Boluda, Perales, et al., 2019).
5 CONCLUSSIONS
This work assessed the performance of DispEn, a new
complexity parameter in EHG characterization to
distinguish term from preterm deliveries in EHG
recordings picked up during regular checkups. Its
performance was compared with the traditionally
used SampEn. Both entropy metrics computed in
window analyses of 120s show statistically
significant differences (p-value < 0.05) in women
who delivered at term from those who delivered
preterm. DispEn outperformed SampEn
discrimination in FWH, unlike WBW, for which
SampEn reached a lower p-value. In both FWH and
WBW the best discrimination was in the 10
th
percentile, with the lowest p-value for DispEn in
FWH and internal parameters c = 7 and m = 2, with
lower entropy values for the preterm group.
ACKNOWLEDGEMENTS
This work was supported by the Spanish ministry of
economy and competitiveness, the European
Regional Development Fund (MCIU/AEI/FEDER,
UE RTI2018-094449-A-I00-AR) and the Generalitat
Valenciana (AICO/2019/220).
REFERENCES
Ahmed, M. U., & Mandic, D. P. (2011). Multivariate
Multiscale Entropy Analysis. IEEE Signal Processing
Letters, 19(2), 91–94. https://doi.org/10.1109/lsp.2011.
2180713
Azami, H., & Escudero, J. (2018). Amplitude- and
fluctuation-based dispersion entropy. Entropy, 20(3),
210. https://doi.org/10.3390/e20030210
Azami, H., Fernández, A., & Escudero, J. (2019).
Multivariate Multiscale Dispersion Entropy of
Biomedical Times Series. Entropy, 21(9), 913.
https://doi.org/10.3390/e21090913
Azami, H., Rostaghi, M., Fernandez, A., & Escudero, J.
(2016). Dispersion entropy for the analysis of resting-
state MEG regularity in Alzheimer’s disease.
Proceedings of the Annual International Conf. of the
IEEE Engineering in Medicine and Biology Society,
EMBS, 2016-Octob, 6417–6420. https://doi.org/10.
1109/EMBC.2016.7592197
Azami, H., Smith, K., & Escudero, J. (2016). MEMD-
enhanced multivariate fuzzy entropy for the evaluation
of complexity in biomedical signals. Proceedings of the
Annual International Conf. of the IEEE Engineering in
Medicine and
Biology Society, EMBS, 2016-Octob,3761-
3764. https://doi.org/10.1109/EMBC.2016.7591546
Bandt, C., & Pompe, B. (2002). Permutation Entropy: A
Natural Complexity Measure for Time Series. Physical
Review Letters, 88(17), 4. https://doi.org/
10.1103/PhysRevLett.88.174102
Benalcazar-Parra, C., Ye-Lin, Y., Garcia-Casado, J.,
Monfort-Ortiz, R., Alberola-Rubio, J., Perales, A., &
Prats-Boluda, G. (2019). Prediction of Labor Induction
Success from the Uterine Electrohysterogram. Journal
of Sensors, 2019, 1–12. https://doi.org/10.1155
/2019/6916251
Berghella, V., Hayes, E., Visintine, J., & Baxter, J. K.
(2008). Fetal fibronectin testing for reducing the risk of
preterm birth. Cochrane Database of Systematic
Reviews 4. https://doi.org/10.1002/14651858.CD00
6843.pub2
de Luca, A., & Termini, S. (1993). A Definition of a
Nonprobabilistic Entropy in the Setting of Fuzzy Sets
Theory. In Readings in Fuzzy Sets for Intelligent
Systems (pp. 197–202). Elsevier. https://doi.org
/10.1016/b978-1-4832-1450-4.50020-1
Devedeux, D., Marque, C., Mansour, S., Germain, G., &
Duchêne, J. (1993). Uterine electromyography: A
critical review. American Journal of Obstetrics and
BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing
266
Gynecology, 169(6), 1636–1653. https://doi.org/10.
1016/0002-9378(93)90456-S
di Marco, L. Y., di Maria, C., Tong, W. C., Taggart, M. J.,
Robson, S. C., & Langley, P. (2014). Recurring patterns
in stationary intervals of abdominal uterine
Electromyograms during gestation. Medical and
Biological Engineering and Computing, 52(8), 707–
716. https://doi.org/10.1007/s11517-014-1174-6
Diaz-Martinez, A., Mas-Cabo, J., Prats-Boluda, G., Garcia-
Casado, J., Cardona-Urrego, K., Monfort-Ortiz, R.,
Lopez-Corral, A., de Arriba-Garcia, M., Perales, A., &
Ye-Lin, Y. (2020). A Comparative Study of Vaginal
Labor and Caesarean Section Postpartum Uterine
Myoelectrical Activity. Sensors, 20(11), 3023. https://
doi.org/10.3390/s20113023
Euliano, T. Y., Nguyen, M. T., Darmanjian, S., Busowski,
J. D., Euliano, N., & Gregg, A. R. (2016). Monitoring
Uterine Activity during Labor: Clinician Interpretation
of Electrohysterography versus Intrauterine Pressure
Catheter and Tocodynamometry. American Journal of
Perinatology, 33(9), 831–838. https://doi.org/10
.1055/s-0036-1572425
Fele-Žorž, G., Kavšek, G., Novak-Antolič, Ž., & Jager, F.
(2008). A comparison of various linear and non-linear
signal processing techniques to separate uterine EMG
records of term and pre-term delivery groups. Medical
and Biological Engineering and Computing, 46(9),
911–922. https://doi.org/10.1007/s11517-008-0350-y
Fuchs, I. B., Henrich, W., Osthues, K., & Dudenhausen, J.
W. (2004). Sonographic cervical length in singleton
pregnancies with intact membranes presenting with
threatened preterm labor. Ultrasound in Obstetrics and
Gynecology, 24(5), 554–557. https://doi.org/10.1002/
uog.1714
Garcia-Casado, J., Ye-Lin, Y., Prats-Boluda, G., Mas-
Cabo, J., Alberola-Rubio, J., & Perales, A. (2018).
Electrohysterography in the diagnosis of preterm birth:
A review. Physiological Measurement, 39(2), 02TR01.
https://doi.org/10.1088/1361-6579/aaad56
Garfield, R. E., & Maner, W. L. (2007). Physiology and
electrical activity of uterine contractions. Seminars in
Cell and Developmental Biology, 18(3), 289–295.
https://doi.org/10.1016/j.semcdb.2007.05.004
Jager, F., Libenšek, S., & Geršak, K. (2018).
Characterization and automatic classification of pre-
term and term uterine records. PLoS ONE, 13(8),
e0202125. https://doi.org/10.1371/journal.pone.0202125
Kafantaris, E., Piper, I., Lo, T. Y. M., & Escudero, J.
(2019). Application of Dispersion Entropy to Healthy
and Pathological Heartbeat ECG Segments.
Proceedings of the Annual International Conf. of the
IEEE Engineering in Medicine and Biology Society,
EMBS, 2269–2272. https://doi.org/10.1109/EMBC.
2019.8856554
Leung, C. (2004). Born too soon. In J. L. CP Howson, MV
Kinney (Ed.), Neuroendocrinology Letters (Vol. 25,
Issue SUPPL 1). https://doi.org/http://whqlibdoc.
who.int/publications/2012/9789241503433_eng.pdf
Mas-Cabo, J., Prats-Boluda, G., Garcia-Casado, J.,
Alberola-Rubio, J., Perales, A., & Ye-Lin, Y. (2019).
Design and Assessment of a Robust and Generalizable
ANN-Based Classifier for the Prediction of Premature
Birth by means of Multichannel Electrohysterographic
Records. Journal of Sensors, 2019, 1–13. https://doi.
org/10.1155/2019/5373810
Mas-Cabo, J., Prats-Boluda, G., Perales, A., Garcia-
Casado, J., Alberola-Rubio, J., & Ye-Lin, Y. (2019).
Uterine electromyography for discrimination of labor
imminence in women with threatened preterm labor
under tocolytic treatment. Medical and Biological
Engineering and Computing, 57(2), 401–411. https://
doi.org/10.1007/s11517-018-1888-y
Mas-Cabo, J., Ye-Lin, Y., Garcia-Casado, J., Díaz-
Martinez, A., Perales-Marin, A., Monfort-Ortiz, R.,
Roca-Prats, A., López-Corral, Á., & Prats-Boluda, G.
(2020). Robust characterization of the uterine
myoelectrical activity in different obstetric scenarios.
Entropy, 22(7), 743. https://doi.org/10.3390/e22070743
Mischi, M., Chen, C., Ignatenko, T., de Lau, H., Ding, B.,
Oei, S. G. G., & Rabotti, C. (2018). Dedicated Entropy
Measures for Early Assessment of Pregnancy
Progression from Single-channel Electrohysterography.
IEEE Transactions on Biomedical Engineering, 65(4),
875–884. https://doi.org/10.1109/TBME.2017.2723933
O’Hara, S., Zelesco, M., & Sun, Z. (2013). Cervical length
for predicting preterm birth and a comparison of
ultrasonic measurement techniques. Australasian
Journal of Ultrasound in Medicine, 16(3), 124–134.
https://doi.org/10.1002/j.2205-0140.2013.tb00100.x
Petrou, S., Yiu, H. H., & Kwon, J. (2019). Economic
consequences of preterm birth: A systematic review of
the recent literature (2009-2017). Archives of Disease
in Childhood, 104(5), 456–465. https://doi.org/
10.1136/archdischild-2018-315778
Pincus, S. M. (1991). Approximate entropy as a measure of
system complexity. Proceedings of the National
Academy of Sciences of the United States of America,
88(6), 2297–2301. https://doi.org/10.1073/pnas.88.6.
2297
Richman, J. S., & Moorman, J. R. (2000). Physiological
time-series analysis using approximate entropy and
sample entropy. American Journal of Physiology-Heart
and Circulatory Physiology, 278(6), H2039–H2049.
https://doi.org/10.1152/ajpheart.2000.278.6.H2039
Rostaghi, M., & Azami, H. (2016). Dispersion Entropy: A
Measure for Time-Series Analysis. IEEE Signal
Processing Letters, 23(5), 610–614. https://
doi.org/10.1109/LSP.2016.2542881
Terrien, J., Marque, C., & Karlsson, B. (2007). Spectral
characterization of human EHG frequency components
based on the extraction and reconstruction of the ridges
in the scalogram. Annual International Conference of
the IEEE Engineering in Medicine and Biology -
Proceedings
, 1872–1875. https://doi.org/10.1109/
IEMBS.2007.4352680
Xiong, J., Liang, X., Zhu, T., Zhao, L., Li, J., & Liu, C.
(2019). A new physically meaningful threshold of
sample entropy for detecting cardiovascular diseases.
Entropy, 21(9), 830. https://doi.org/10.3390/e210
90830.
Dispersion Entropy: A Measure of Electrohysterographic Complexity for Preterm Labor Discrimination
267