Content uploaded by Kemal Hansu
Author content
All content in this area was uploaded by Kemal Hansu on Apr 17, 2024
Content may be subject to copyright.
383
© 2024 Nigerian Journal of Clinical Practice | Published by Wolters Kluwer ‑ Medknow
Background: Every day, 810 women die of preventable causes related to
pregnancy and childbirth worldwide, and preeclampsia is among the top three
causes of maternal deaths. Aim: To develop a diagnostic system with articial
intelligence for the early diagnosis of preeclampsia. Methods: This retrospective
study included pregnant women who were screened for the inclusion criteria on
the hospital’s database, and the sample consisted of the data of 1158 pregnant
women diagnosed with preeclampsia and 9194 pregnant women who were
not diagnosed with preeclampsia at Kahramanmaras Necip Fazıl City Hospital
Gynecology and Pediatrics Additional Service Building, Kahramanmaras/Turkey.
The statistical analysis was performed using the Statistical Package for social
sciences (SPSS) version 22 for windows. Articial intelligence models were
created using Python, scikit-learn, and TensorFlow. Results: The model achieved
73.7% sensitivity (95% condence interval (CI): 70.2%–77.1%) and 92.7%
specicity (95% CI: 91.7%–93.6%) on the test set. Furthermore, the model had
90.6% accuracy (95% CI: 90.1% - 91.1%) and an area under the curve (AUC)
value of 0.832 (95% CI: 0.818‑0.846). The signicant parameters in predicting
preeclampsia in the model were hemoglobin (HGB), age, aspartate transaminase
level (AST), alanine transferase level (ALT), and the blood group. Conclusion:
Articial intelligence is eective in the prediction and diagnosis of preeclampsia.
Articial intelligence, diagnostic method, preeclampsia, pregnancy
A Bülez, K Hansu1, ES Çağan2, AR Şahin3, HÖ Dokumacı4
Address for correspondence: Dr. A Bülez,
Kahramanmaras Sütcü Imam University, Turkey.
E‑mail: aysel.bulezz@gmail.com
to maternal health, it should be detected in the early
period and managed appropriately to reduce maternal
morbidity and mortality rates.[3] Today, the etiology of
preeclampsia is not clear, and there is still no single
eective treatment for this condition.
The diagnostic criteria for preeclampsia are dened
by the American College of Obstetricians and
Gynecologists (ACOG) in accordance with the clinical
variability, pathogenesis, multisystemic eects, and
prognostic markers of preeclampsia.[6] Accordingly, if a
previously normotensive pregnant woman has a systolic
blood pressure (SBP) ≥140 mmHg or diastolic blood
pressure (DBP) ≥90 mmHg in two measurements taken
at least four hours apart after the 20th gestational week
Original Article
Hypertensive disorders of pregnancy, including
preeclampsia, cause various complications in
approximately 10% of pregnancies worldwide, with
an increased incidence of 25% over the past two
decades.[1] Preeclampsia aects 4.6% of pregnant women.
Fetal growth is impaired in preeclamptic pregnant
women, leading to low birth weight as a predisposing
factor to neonatal deaths.[2] Every day, 810 women die
of preventable causes related to pregnancy and childbirth
worldwide, and preeclampsia is among the top three
causes of maternal deaths.[3,4] Within the scope of the
global “Sustainable Development Goals,” it is aimed to
reduce the global maternal mortality rate to below 70 per
100,000 births by 2030.[5] In this context, the prevention
of maternal deaths from preventable causes and the early
diagnosis and treatment of risky conditions are important.
Considering the threat of preeclampsia and its connection
Departments of Midwifery,
1Gynecology and Obstetrics
and 4Electrical and Electronic
Engineering, Kahramanmaras
Sutcu Imam University,
2Department of Midwifery,
Agri Ibrahim Cecen
University, 3Department
of Infectious Diseases
and Clinic Microbiology,
University of Health
Sciences, Adana City Health
Research Center, Turkey
How to cite this article: Bülez A, Hansu K, Çağan ES, Şahin AR,
Dokumacı HÖ. Artificial Intelligence in Early Diagnosis of Preeclampsia.
Niger J Clin Pract 2024;27:383‑8.
Received:
28-Mar-2023;
Revision:
19-Jan-2024;
Accepted:
08-Feb-2024;
Published:
26-Mar-2024
This is an open access journal, and arcles are distributed under the terms of the
Creave Commons Aribuon‑NonCommercial‑ShareAlike 4.0 License, which allows
others to remix, tweak, and build upon the work non‑commercially, as long as
appropriate credit is given and the new creaons are licensed under the idencal
terms.
For reprints contact: WKHLRPMedknow_reprints@wolterskluwer.com
Access this article online
Quick Response Code:
Website: www.njcponline.com
DOI: 10.4103/njcp.njcp_222_23
PMID: *******
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
Bülez, et al.: Preeclampsia and articial intelligence
384 Nigerian Journal of Clinical Practice ¦ Volume 27 ¦ Issue 3 ¦ March 2024
with accompanying proteinuria, she is diagnosed with
preeclampsia.[6,7] The diagnostic criteria of ACOG, which
include the diagnostic criteria for preeclampsia, except
for hypertension and proteinuria, are thrombocytopenia,
renal failure, deteriorated liver enzymes, and cerebral
and visual disorders.[6] Edema, which is among the signs
of preeclampsia, can be encountered in the majority of
healthy pregnant women. For this reason, edema was
removed from the diagnostic criteria by the National
Heart, Lung, and Blood Institute (NHLBI) working
group.[7] Preeclampsia cases are classied as early onset
preeclampsia (EOP) if they develop before 34 weeks
of gestation or late-onset preeclampsia (LOP) if they
develop at or after 34 weeks of gestation.[6]
The delivery of the fetus is recommended as a treatment
method for preeclampsia, but current studies on new
treatments are promising. Management consists of
preconception counseling, perinatal blood pressure
control, the management of complications, the timely
delivery of the fetus, and postnatal monitoring.
ACOG recommends preconception counseling for any
woman who has previously experienced preeclampsia.
Prenatal management is recommended for women with
preeclampsia excluding those with severe symptoms
before the 37th week of pregnancy. After 37 weeks,
delivery is recommended instead of observation. Maternal
stabilization and delivery are recommended for women
with preeclampsia accompanied by severe symptoms at
34 weeks or later or women showing unstable maternal
or fetal conditions regardless of gestational age.[6,8] It is
particularly important to identify pregnant women at a
high risk of preeclampsia in the rst trimester. Developing
a prediction model for pregnant women is important in
terms of providing early intervention and improving
maternal and newborn outcomes. Thus, such a model
can increase the possibility to identify pregnant women
with a high risk of preeclampsia.[9] Over the past two
decades, many researchers have created various predictive
models for preeclampsia (general health condition of the
pregnant woman, serum biochemical indicators, Doppler
ultrasound, and mean arterial pressure).[9] However, today,
the absence of screening tests that can reliably detect
risks in the early stage of pregnancy and evidence-based
denitive strategies to prevent preeclampsia limits the
eective management of preeclampsia. For this reason,
the diagnosis of preeclampsia can be made after the
20th gestational week and only after symptom formation.[7]
Articial intelligence (AI) has been a topic of interest
among researchers and biomedical industries because of
its ability to process large sets of data, produce accurate
results, and control processes to achieve the most
optimized result.[10] As in many other elds, AI has been
widely used in the eld of health for predictive modeling,
diagnosis, early diagnosis, and monitoring.[11,12] Because
of the impact and benets of the increase in the use of
AI in healthcare services, its use in the early diagnosis of
preeclampsia is expected to increase. In line with studies
performed so far, some approaches including metabolite
measurements, image analyses, and risk factor data sets
have been used to diagnose and predict preeclampsia,
among other diagnostic methods, within the scope of
AI applications.[13] It was stated that AI applications
are quite eective in the diagnosis of preeclampsia.[2]
Liu et al.[9] (2022) reported that the AI application they
developed in their study was a promising tool for the
early diagnosis of preeclampsia. Schmidt et al.[14] (2022)
found that AI techniques constituted a valid approach
to improving the prediction of adverse outcomes in
pregnant women at a high risk of preeclampsia compared
with current standard clinical techniques. There is no AI
tool for the early diagnosis of preeclampsia in Turkey.
This study aimed to develop an AI application for the
early diagnosis and treatment of preeclampsia in Turkey.
Study design and ethical approval
This study was retrospective in design. The ethical
approval was obtained from the local ethics
committee (Kahramanmaras Medical Faculty,
protocol decision dated 23.11.2021 and numbered 16,
Kahramanmaras/Turkey). In addition, written permission
was obtained from the institution where the study was
conducted. The study was conducted in accordance with
the principles of the Declaration of Helsinki.
Population and sample
The hospital records of 37,823 pregnant women, who
presented to Kahramanmaras Necip Fazıl City Hospital
Gynecology and Pediatrics Additional Service Building in
the Kahramanmaras/Turkey between January 1, 2018, and
November 18, 2021, were screened based on the inclusion
criteria. Among these cases, there were 1732 cases
diagnosed with preeclampsia, based on their ICD‑10 codes.
The remaining 37,828 cases did not have the diagnosis of
preeclampsia in their les. A sample selection method was
not used in the study. The archival data of pregnant women
who met the inclusion criteria constituted the sample.
Although archival data were available for 1732
preeclamptic pregnant women, the sample excluded the
data of 70 women with recurrent records, 219 whose
preliminary information was not available, and 285 who
had incomplete information about laboratory ndings.
Therefore, the sample included the data of n: 1158 women
diagnosed with preeclampsia. Furthermore, although
archival data were available for 37,823 pregnant women
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
Bülez, et al.: Preeclampsia and articial intelligence
385
Nigerian Journal of Clinical Practice ¦ Volume 27 ¦ Issue 3 ¦ March 2024
without the diagnosis of preeclampsia, the sample excluded
the data of 27,748 women with incomplete information
about their laboratory ndings and 881 whose medical
histories were unavailable. Therefore, the sample included
the data of n: 9194 pregnant women who were not
diagnosed with preeclampsia. The total number of women
whose data were included in the study was n: 10,307.
Inclusion and exclusion criteria
Inclusion criteria for the case group
The inclusion criteria for the case group were meeting
the ACOG diagnostic criteria for preeclampsia,
SBP ≥140 mmHg or DBP ≥90 mmHg in two
measurements taken at least four hours apart after
20 weeks of gestation in a previously normotensive
pregnant woman, proteinuria (urine protein ≥300 mg
in 24‑hour urine or protein/creatinine ≥0.3 mg/dl),
multiple positive results for proteinuria in urine protein
dipstick test measurements at dierent times, and being
a pregnant woman between the ages of 18 and 45 years.
Inclusion criteria for the control group
The inclusion criteria for the control group were
the exclusion of the ACOG diagnostic criteria for
preeclampsia and being a pregnant woman between the
ages of 18 and 45 years.
Exclusion criteria
Being younger than 18 years or older than 45 years, having
a chronic disease such as chronic kidney disease, type 1
diabetes, or systemic lupus erythematosus, and incomplete
or insucient le data were the exclusion criteria.
Data collection
After obtaining the permission of the institution,
patient data were extracted from the database under the
supervision of the data processing ocer of the hospital.
Information was collected about sociodemographic
characteristics, anthropometric measurements, vital
signs, obstetric features, laboratory ndings, presence
of chronic diseases, lifestyle characteristics, and family
history (history of chronic diseases in the family).
Clinical information included patient age, blood type,
HGB, AST, ALT, hyperemesis status, blood pressure,
diabetes status, heart rate, hypertrophy status, anemia
status, headache status, metabolic conditions, and PCOS
diagnosis status. Some parameters could not be included
in the analyses because they were not registered in the
system. These parameters included paternal blood type,
stillbirth history, miscarriage history, height, and some
laboratory data (homocysteine, cholesterol, triglyceride,
and ferritin). Data collected from the records of the
preeclampsia-positive and -negative patients were
randomly divided into training, validation, and test
sets. The test set was used to test the predictions of the
machine learning model on the data that the model had
not seen during training.
Data analysis
The descriptive characteristics of the patients (e.g. age,
number of pregnancies, and number of live births)
were analyzed using the SPSS 22 package program.
Descriptive statistics, including mean, standard
deviation, frequency, and percentage values, were
calculated. In the AR testing phase, the data were
coded, and an AI diagnostic system was created. The
AI models were created using Python, scikit-learn, and
TensorFlow. The diagnostic power of the AI models was
evaluated with ROC curve, specicity, and sensitivity
analyses.
Machine learning
Gradient boosting is a machine learning technique
that produces a prediction model in the form of an
ensemble of weak prediction models, typically decision
trees.[15,16] It builds the model in a series fashion by
allowing for the optimization of a dierentiable loss
function. Gradient boosting decision tree (GBDT) is
a widely used machine learning algorithm thanks to
its eciency, accuracy, and interpretability. GBDT
achieves state-of-the-art performance levels in many
machine learning tasks. LightGBM is a highly ecient
GBDT implementation,[17] and it was used throughout
this study. The Python implementation of LightGBM
was used in the study. The augmented decision tree
regression method was used as the methodology.
Twenty percent of the data were allocated as a test set.
The models were trained by 5-fold cross-validation on
the remaining data. Hyperparameter optimization was
performed with the grid search of several LightGBM
parameters. Finally, the models were evaluated on the
holdout test set.
Sociodemographic characteristics of patients
The mean age of the patients who were not diagnosed
with preeclampsia was 27.61 ± 5.99 (min: 18, max:
47), the mean number of their pregnancies was
2.85 ± 1.51 (min: 1, max: 14), and their mean total
number of childbirths was 1.84 ± 1.18 (min: 0, max: 13).
The mean age of the patients who were diagnosed with
preeclampsia was 29.01 ± 7.04 (min: 18, max: 47), the
mean number of their pregnancies was 2.52 ± 1.33 (min:
1, max: 6), and their mean total number of childbirths
was 1.35 ± 0.75 (min: 1, max: 6) [Table 1].
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
Bülez, et al.: Preeclampsia and articial intelligence
386 Nigerian Journal of Clinical Practice ¦ Volume 27 ¦ Issue 3 ¦ March 2024
Model test results
The model achieved 73.7% sensitivity (95% condence
interval (CI): 70.2%–77.1%) and 92.7% specicity (95%
CI: 91.7%–93.6%) on the test set. Furthermore, the
model had 90.6% accuracy (95% CI: 90.1%–91.1%) and
an AUC value of 0.832 (95% CI: 0.818-0.846). Based
on these results, the model did a good job predicting
preeclampsia based on the characteristics of the patients,
especially adverse parameters.The results of the model
tests are shown in Table 2.
Figure 1 shows the signicance levels of the tested
characteristics in the model. The most signicant
characteristics were HGB, age, AST, ALT, and blood
type, respectively.
Preeclampsia is a complex pregnancy-related condition
that can lead to signicant fetal‑maternal morbidity or
even mortality.[13] Therefore, the early identication
of women who are at risk of preeclampsia is of great
importance.[18] Multiple screening strategies have been
developed over time to achieve the best results for
the early diagnosis of preeclampsia.[19] AI is among
the systems that have been developed in recent years.
This study aimed to develop an AI application for
the early diagnosis and treatment of preeclampsia
in Turkey. In the study, the sensitivity of the AI
model was found to be 73.7% (95% condence
interval (CI): 70.2%–77.1%), whereas its specicity
was 92.7% (95% CI: 91.7%–93.6%) in the test set.
Moreover, the accuracy of the model was determined
to be 90.6% (95% CI: 90.1%–91.1%), and its AUC
value was 0.832 (95% CI: 0.818-0.846). According to
the model, the most signicant characteristics in the
early diagnosis of preeclampsia were HGB, age, AST,
ALT, and blood type, respectively. In their retrospective
medical record review study, Liu et al.[9] (2022)
used 5 machine learning algorithms (deep neural
network (DNN), logistic regression (LR), support
vector machine (SVM), decision tree (DT), and random
forest (RF)) to diagnose preeclampsia early, and
they examined 18 variables in the model, including
maternal characteristics, medical history details,
prenatal laboratory results, and ultrasound results. The
authors stated that the methods they used automatically
identied a number of signicant predictive
characteristics in the early diagnosis of preeclampsia
and produced high predictive success (based on clinical
history and prenatal screening results) in the assessment
of risk based on early pregnancy information.[9]
Sufriyana et al.[2] (2020) compared six machine learning
models in their study on the diagnosis of preeclampsia,
the best model was a random forest algorithm
consisting of 500 trees, and the model contained
mostly data from the medical backgrounds of patients.
The researchers argued that the model they applied
outperformed previously developed models, especially
in terms of accuracy, using maternal characteristics
and medical history details without biophysical or
biochemical markers.[2] Schmidt et al. (2022) applied
two models to improve the prediction of adverse
outcomes associated with preeclampsia and reported
that the random forest classier performed slightly
less well than other available metrics. Compared with
current clinical standard techniques, machine learning
techniques are a valid approach to improving the
prediction of adverse outcomes in pregnant women
at a high risk of preeclampsia.[14] In the study, they
Accuracy 90.6 91.2 90.5 90.8 90.1
Sensitivity 70.2 72.4 72.9 75.6 77.3
Specicity 93.1 93.5 92.6 92.7 91.7
AUC 0.817 0.830 0.827 0.841 0.845
Preeclamptic
(n
(n
Age 29±7.1 27.6±6
HGB 10.5±1.6 11.6±1.4
AST 23±18 15.8±7.3
ALT 13.3±13.5 12.6±10.2
Hyperemesis gravidarum 5 (0.4%) 245 (2.7%)
Hypertension 11 (1%) 22 (0.2%)
Diabetes 10 (0.9%) 58 (0.6%)
Heart disease 0 (0%) 8 (0.1%)
Hyperthyroidism 35 (3.1%) 220 (2.4%)
Kidney disease 509 (45.2%) 4575 (49.8%)
Anemia 1019 (90.4%) 8525 (92.7%)
Headache 399 (35.4%) 3079 (33.5%)
Metabolic disease 25 (2.2%) 278 (3%)
PCOS 6 (0.5%) 31 (0.3%)
Feature importance levels in the LightGBM model
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
Bülez, et al.: Preeclampsia and articial intelligence
387
Nigerian Journal of Clinical Practice ¦ Volume 27 ¦ Issue 3 ¦ March 2024
performed on the data of 11,006 pregnant women, Jhee
et al.[20] (2019) revealed that systolic blood pressure,
serum blood urea nitrogen and creatinine levels, platelet
counts, serum potassium levels, white blood cell
counts, serum calcium levels, and urine protein were
the most signicant variables in the prediction models
they used for the early diagnosis of preeclampsia, and
they claimed that the combined use of maternal factors
and common antenatal laboratory data from the early
second trimester to the third trimester can eectively
predict late-onset preeclampsia using machine learning
algorithms. In their study that examined the data of
16,370 pregnant women, Marić et al.[21] (2020) used
two models, the elastic mesh and gradient boosting
algorithms, there were 67 variables taken into account
in the models (e.g., maternal characteristics, medical
history, routine prenatal laboratory results, and drug
intake), and they stated that the models automatically
identied a number of signicant characteristics for
the early diagnosis of preeclampsia and showed high
predictive performance for preeclampsia risk based
on routine early pregnancy information. In their
prospective case-control study in Romania, Melinte
et al.[19] (2022) implemented 4 models for the diagnosis
of preeclampsia (decision tree (DT), naive Bayes (NB),
support vector machine (SVM), and random forest),
they reported that the models showed high performance
in the early diagnosis of preeclampsia, and they asserted
that models based on machine learning can be useful
tools for the prediction of preeclampsia in the rst
trimester of pregnancy. Ethnicity could not be taken
into account in this study because of the lack of records.
In the thesis study by Bennett et al. (2021) in which
an algorithmic modication of Deep Neural Networks
(DNN) was developed to identify high-risk patients in
the diagnosis of preeclampsia, the authors stated that
patient race/ethnicity should be taken into account more
thoroughly in the prediction of preeclampsia.[22]
Limitations
The limitations of this study, which was conducted as a
retrospective study, included the lack of some records,
duplications in the data recording system, dierences
in routine tests performed on pregnant women, the lack
of records on some sociodemographic characteristics
of pregnant women (e.g., family history and lifestyle
habits), and the inaccessibility of some data because the
data recording system was kept and archived in physical
les instead of an online recording system.
Articial intelligence using the decision tree regression
technique was an eective predictive method in the
early diagnosis of preeclampsia. With this method, it
will be possible to establish early warning systems by
standardizing the data recording systems of hospitals,
transferring these records to the computer environment
regularly and completely, and conducting new studies on
other health problems with the same methods.
Considering the importance of the early diagnosis of
preeclampsia, it is recommended to adapt AI applications
to all clinical environments and inform healthcare
personnel about AI applications. Future studies should
consider implementation of this diagnostic tool in health
institutions to improve maternal and infant health and
prevent possible risks associated with the disease.
Ethics approval and consent to participate
Ethical approval was obtained from the local ethics
committee (Kahramanmaras Medical Faculty in the
Kahramanmaras/Turkey, protocol decision dated
23.11.2021 and numbered 16). In addition, written
permission was obtained from the institution where
the study was conducted. The study was conducted in
accordance with the principles of the Declaration of
Helsinki.
Acknowledgment
The authors would like to thank all hospital sta who
consented to take part in the conduct of the study.
Financial support and sponsorship
Nil.
Conicts of interest
There are no conicts of interest.
1. Nakhleh MK, Baram S, Jeries R, Salim R, Haick H, Hakim M.
Articially intelligent nanoarray for the detection of preeclampsia
under real-world clinical conditions. Adv Mater Technol
2016;1:1600132.
2. Sufriyana H, Wu YW, Su EC. Articial intelligence‑assisted
prediction of preeclampsia: Development and external validation
of a nationwide health insurance dataset of the BPJS Kesehatan
in Indonesia. EBioMedicine 2020;54,102710.
3. World Health Organization (WHO). Maternal Mortality.
Available from: https://www.who.int/news-room/fact-sheets/
detail/maternal-mortality. [Last accessed on 2022 Dec 25].
4. World Health Organization (WHO). New global targets to
prevent maternal deaths. Available from: https://www.who.int/
news/item/05-10-2021-new-global-targets-to-prevent-maternal-
deaths. [Last accessed on 2021 Nov 11].
5. United Nations Development Programme (UNDP). Sustainable
development goals. Available from: https://www.kureselamaclar.
org/amaclar/saglik-ve-kaliteli-yasam/. [Last accessed on
2022 Dec 25].
6. American College of Obstetricians and Gynecologists (ACOG).
Hypertension in pregnancy. Report of the American College of
Obstetricians and Gynecologists’ Task Force on Hypertension in
Pregnancy. Obstet Gynecol 2013;122:1122‑31.
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
Bülez, et al.: Preeclampsia and articial intelligence
388 Nigerian Journal of Clinical Practice ¦ Volume 27 ¦ Issue 3 ¦ March 2024
7. Akalın A, Şahin N. Preeklampsi: Current approaches in diagnosis
and nursing management. J Hum Rhythm 2018;4:88‑97.
8. Phipps EA, Thadhani R, Benzing T, Karumanchi SA.
Pre-eclampsia: Pathogenesis, novel diagnostics and therapies. Nat
Rev Nephrol 2019;15:275‑89.
9. Liu M, Yang X, Chen G, Ding Y, Shi M, Sun L, et al.
Development of a prediction model on preeclampsia using
machine learning-based method: A retrospective cohort study in
China. Front Physiol 2022;13:896969.
10. Manickam P, Mariappan SA, Murugesan SM, Hansda S,
Kaushik A, Shinde R, et al. Articial Intelligence (AI) and
internet of medical things (IoMT) assisted biomedical systems
for intelligent healthcare. Biosensors 2022;12:562.
11. Ramakrishnan R, Rao S, He JR. Perinatal health
predictors using articial intelligence: A review. Women’s
Health 2021;17,17455065211046132. doi:
10.1177/17455065211046132.
12. Can B, Başer A, Baktır AS, Özceylan G, Kolcu G. Articial
intelligence in health. Educ Med J SDU 2021;28:355‑9.
13. Martínez‑Velasco A, Martínez‑Villaseñor L, Miralles‑Pechuán L.
Machine learning approach for pre-eclampsia risk factors association.
In: Proceedings of the 4th EAI International Conference on Smart
Objects and Technologies for Social Good. 2018. p. 232-7.
14. Schmidt LJ, Rieger O, Neznansky M, Hackelöer M, Dröge LA,
Henrich W, et al. Machine-learning–based algorithm improves
prediction of preeclampsia-associated adverse outcomes. Am J
Obstet Gynecol 2022;227:77.e1‑30.
15. Jerome HF. Greedy function approximation: A gradient boosting
machine. Ann Statist 2001;29:1189‑232.
16. Jerome HF. Stochastic gradient boosting. Comput Statist Data
Anal 2002;38:367‑78.
17. Ke G, Meng Q, Finley T, Wang T, Chen W, Ma W, et al.
LightGBM: A highly ecient gradient boosting decision tree.
Adv Neural Inf Process Syst 2017;3148,3156.
18. Hackelöer M, Schmidt L, Verlohren S. New advances in
prediction and surveillance of preeclampsia: Role of machine
learning approaches and remote monitoring. Arch Gynecol
Obstet 2023;308:1663‑77.
19. Melinte‑Popescu A‑S, Vasilache I‑A, Socolov D,
Melinte-Popescu M. Predictive performance of machine
learning-based methods for the prediction of preeclampsia-a
prospective study. J Clin Med 2023;12:418.
20. Jhee JH, Lee S, Park Y, Lee SE, Kim YA, Kang SW, et al.
Prediction model development of late-onset preeclampsia using
machine learning‑based methods. PLoS One 2019;14:e0221202.
21. Marić I, Tsur A, Aghaeepour N, Montanari A, Stevenson DK,
Shaw GM, et al. Early prediction of preeclampsia via machine
learning. Am J Obstet Gynecol 2020;2:100100.
22. Bennett R, Mulla ZD, Parikh P, Hauspurg A, Razzaghi T. An
imbalance-aware deep neural network for early prediction of
preeclampsia. Plos one, 2022:17;e0266042.
Downloaded from http://journals.lww.com/njcp by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AW
nYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 04/17/2024
