You are currently viewing a new version of our website. To view the old version click .
MDPIMDPI
Search all articles
Big Data and Cognitive Computing
Download PDF
  • Review
  • Open Access

9 February 2023

Prediction of Preeclampsia Using Machine Learning and Deep Learning Models: A Review

,
,
,
,
,
,
,
,
and
1
Department of Computer Science, College of Computer Science and Information Technology, Imam Abdulrahman Bin Faisal University, P.O. Box 1982, Dammam 31441, Saudi Arabia
2
Department of Computer Engineering, College of Computer Science and Information Technology, Imam Abdulrahman Bin Faisal University, P.O. Box 1982, Dammam 31441, Saudi Arabia
3
Computer Science Department, Applied College, Imam Abdulrahman Bin Faisal University, P.O. Box 1982, Dammam 31441, Saudi Arabia
*
Author to whom correspondence should be addressed.
Big Data Cogn. Comput.2023, 7(1), 32;https://doi.org/10.3390/bdcc7010032
Version Notes
Order Reprints

Abstract

Preeclampsia is one of the illnesses associated with placental dysfunction and pregnancy-induced hypertension, which appears after the first 20 weeks of pregnancy and is marked by proteinuria and hypertension. It can affect pregnant women and limit fetal growth, resulting in low birth weights, a risk factor for neonatal mortality. Approximately 10% of pregnancies worldwide are affected by hypertensive disorders during pregnancy. In this review, we discuss the machine learning and deep learning methods for preeclampsia prediction that were published between 2018 and 2022. Many models have been created using a variety of data types, including demographic and clinical data. We determined the techniques that successfully predicted preeclampsia. The methods that were used the most are random forest, support vector machine, and artificial neural network (ANN). In addition, the prospects and challenges in preeclampsia prediction are discussed to boost the research on artificial intelligence systems, allowing academics and practitioners to improve their methods and advance automated prediction.

1. Introduction

Placental dysfunction-related disorders (PDDs), such as preeclampsia and intrauterine growth restriction, require that a referral choice be made quickly. Preeclampsia is the world’s most common cause of maternal death and morbidity [1]. It is a hazardous medical condition that can develop about halfway through gestation (after 20 weeks) and is associated with substantial mortality and morbidity for the mother, the fetus, and the newborn [2]. It is a pregnancy complication that affects from 3% to 7% of all pregnant women, whether in their first or subsequent pregnancies, and is identified by new-onset proteinuria and gestational hypertension, usually by the last trimester of pregnancy [3]. The disease impacts mothers and limits fetal growth, resulting in low birth weights, a risk factor for neonatal mortality [4]. The symptoms of preeclampsia include swelling, blurred vision, headaches, high blood pressure, protein in the urine [2], and stress on the mother’s heart and other organs. It also affects the mother’s blood supply to the placenta, weakens kidney and liver functions, causes fluid build-up in the lungs, and causes other serious complications [5].
Preeclampsia occurs in approximately 5.37 out of every 10,000 women in Saudi Arabia [2]. It can result in unfavorable pregnancy outcomes, such as neurological consequences for the newborn. The pathogenesis and etiology of preeclampsia are still unknown, and delivery is the only possible treatment for a pregnant woman diagnosed with preeclampsia. However, to avoid difficulties and improve outcomes, it is critical to detect preeclampsia risk before pregnancy [6].
Artificial intelligence (AI) is the study of concepts that can be used to build machines capable of thinking, judging, and intending in accordance with standard human reactions to stimulation. Systems that integrate powerful software, hardware, knowledge-based processing models, and extensive databases to simulate the characteristics of efficient human decision making are defined as AI. AI is employed in many industries, including scientific research, medical prognosis, robot control, and law [7].
AI in the medical field may be divided into virtual and physical categories. The physical category focuses on assisted surgical robots, intelligent prosthetics for the disabled, and geriatric care. The virtual component includes tools such as a neural network-based treatment decision support and electronic health record systems. It also includes machine learning (ML), which relies on mathematical algorithms that enhance learning via experience [8].
ML is crucial for helping the model learn and adjust based on the input data without being explicitly programmed. It is the concept of providing machines with the capability to learn and understand data, recognize patterns, and make predictions or decisions [9]. Therefore, ML can be trained to identify patterns in the same way that doctors do. It can help diagnose or predict diseases, recognize patient risk factors, and promote the research and development of new drugs. It is beneficial in situations in which the diagnostic data a doctor looks at has already been digitized. Such situations include using computerized tomography scans to detect lung cancer or strokes, analyzing cardiac magnetic resonance images and electrocardiograms to determine the likelihood of sudden cardiac death or other heart conditions, examining eye images for signs of diabetic retinopathy, and classifying skin lesions on the basis of skin imaging data. An example of success in these fields is a study by Bhatia et al. [10], which proposed a method for detecting lung cancer using deep residual networks for feature extraction, achieving an accuracy of 84%.
Data are the basis for ML models, and, when high-quality data are abundant, algorithms can diagnose on par with professionals. The key differences are the algorithms’ ability to make conclusions in a split second and their ability to be readily replicated anywhere around the globe. The aim is that all people, wherever they are, can access affordable and top-quality diagnostic services.
Deep learning (DL) is a form of AI that can tackle complex problems that may be challenging or even impossible for traditional AI techniques to solve. One of the key benefits of DL is its ability to utilize both labeled and unlabeled data during training, which enables it to effectively handle diverse information and learn from it. Additionally, DL is well suited for working with large datasets; therefore, its applications are likely to expand in the future. Many recent studies have demonstrated the capabilities of DL technologies, including the ability to learn from complex data, perform image recognition, and categorize text [11]. In a study by Tahir et al. [12], the researchers aimed to use a neural network (NN) [13] to estimate the probability of preeclampsia and compared its performance to other algorithms such as naïve Bayes (NB) and linear regression. They also tested the NN with one hidden layer and found that using 17 neurons resulted in the lowest error rate. The model was then validated using three different methods and was found to have the best performance, with an accuracy of 96.66%, when validated using the leave-one-out (LOO) cross-validation method.
This paper provides an in-depth review of the most recent studies on several preeclampsia prediction methods that use clinical data and employ DL and ML. Twenty-five articles released since 2018 are tabulated, grouped, and analyzed from many angles, including ML and DL models, dataset size, and performance. Key search terms such as “preeclampsia”, “artificial intelligence”, “machine learning”, and “deep learning” were used to identify relevant studies for inclusion in this review. The primary aim of this review is to provide a comprehensive overview of the current state of research in the realm of automated preeclampsia prediction. Furthermore, this review aims to delve into the challenges and opportunities in the field of preeclampsia research.
The remainder of this work is organized as follows: Section 2 presents the studies that used statistical, ML, and DL methods to predict preeclampsia. Section 3 discusses the most widely used algorithms and data types. Section 4 presents the challenges and opportunities in preeclampsia prediction. Lastly, Section 5 provides the conclusion.

3. Discussion

This review assessed the latest research on general methods, ML, and DL for preeclampsia prediction. Our goal was to define the data types and techniques that were employed in preeclampsia prediction, as well as the methods that delivered meaningful outcomes. In this section, the data used for the prediction of preeclampsia, the algorithms used, and the limitations of the reviewed studies are discussed.
Examination of the studies revealed that preeclampsia was predicted using a variety of data sources, including clinical data, questionnaires, and laboratory test data. Most of the studies that predicted preeclampsia with high accuracy used clinical information, such as maternal and gestational age, blood pressure, height, weight, and history of preeclampsia and hypertension. However, the best studies used mean arterial pressure, proteinuria, creatinine, and PI. Jhee et al. [19] used the combination of maternal factors and common antenatal laboratory data from the early second and third trimesters, which helped in effectively predicting late-onset preeclampsia. Modak et al. [17] combined UPCR and a UA Doppler screening test to predict preeclampsia and produced results with high accuracy. Serra et al. [15] combined the maternal characteristics, biophysical parameters, and PlGF for the screening of eoPE. Similarly, Schmidt et al. [26] predicted preeclampsia with high accuracy using the ratio of sFlt1 to PlGF.
Nevertheless, it should be emphasized that the studies in this review that had an accuracy above 90% suffer from several limitations. For example, Jhee et al. [19] developed an SGB model for late-onset preeclampsia prediction that achieved an accuracy of 0.973. A model for first-trimester data could not be obtained because most of the pregnant women were only involved in the antenatal evaluation program in the early second trimester. Despite this lack of data in the first trimester, the developed model predictive power was sufficient. Furthermore, the number of preeclampsia incidents was small, as only 474 out of 10,532 cases had preeclampsia, which is a common limitation from which published models suffer.
Tahir et al. [32] implemented NN to estimate the probability of preeclampsia with 96.66% accuracy after being validated using the LOO cross-validation method. The study used a sample of records from 239 women and considered 17 features. The dataset size is considered small; consequently, further research in a large cohort is essential to prove the reliability of the model result and minimize the features needed. In addition, there is no indication of whether the model is used for early or late prediction of preeclampsia. Another study by Tahir et al. [12] implemented a preeclampsia prediction DL model with an accuracy of 95.68% with a PSO feature selection algorithm to reduce the number of features from 17. The dataset was gathered from 1077 patient records at two hospitals in Makassar, as well as Surabaya Haji General Hospital, including records from the same 239 women as were included in the previous study. The PSO feature selection algorithm reduced the number of features from 17 to nine, which enhanced the dataset’s quality and improved the learning process; however, as in the previous study, early- and late-onset preeclampsia were not differentiated.
Furthermore, in Serra et al.’s study [15], despite the high accuracy of the mathematical model, it also suffers from the low number of eoPE events, which comprised only 17 of the 6893 pregnancies, causing an imbalanced database. Correspondingly, the prospective observational study by Modak et al. [17] produced an accuracy of 92.24%, but suffered from dataset-related problems. The study was conducted in a sample of 116 pregnant mothers, which is considered small; thus, to validate the reliability of the method, further study in a large cohort is required.
Carreno et al. [23] used a comparative strategy to assess the utility of time-series summary methods and feature size reduction methods in preeclampsia prognosis. The highest accuracy was achieved with an SVM algorithm paired with SPD for feature clustering, peaking at 93%. However, in this study, preeclampsia referred to both early and late preeclampsia, which makes it difficult to evaluate the proposed method’s performance in further investigations because patients could have early, late, or no preeclampsia.
Additionally, Li et al. [22] developed an XGBoost preeclampsia prediction model with an accuracy of 0.920. However, the performance of the XGBoost model could not be quantified among pregnancies with eoPE because of the rarity of eoPE. Therefore, additional research is needed to build more prediction models for the prediction of eoPE. Another limitation is that the model did not include the features related to pregnancy-associated plasma protein A or PlGF, which have previously been demonstrated to be related to the incidence of preeclampsia [37].
Moreover, research conducted by Sufriyana et al. [27] developed a CVR model for the prediction of intrauterine growth restriction (IUGR) and preeclampsia, including a sample of 95 women and considering 13 features. The CVR achieved an accuracy of 90.6%; however, the model has several limitations. The first limitation is that the model does not differentiate between IUGR and preeclampsia; consequently, the model should only be used for a referral decision. Second, the CVR model cannot be used to decide whether to deliver before term, as such a decision must be made based on models that precisely predict severe incidents of IUGR and preterm or early-onset preeclampsia. Third, the study used a small dataset of only 95 pregnancies.
Sakinah et al. [33] built an LSTM model for preeclampsia prediction with an accuracy of 90.22%. However, more information about the study needs to be provided. The first missing information is the dataset size used in the model. The other information that needs to be mentioned is whether the prediction is for early- or late-onset preeclampsia. Figure 1 illustrates the ML and DL algorithms that were most widely implemented in the studies, Figure 2 shows the algorithms that obtained the best results in each study, and Table 4 shows a summary of the previous studies that used clinical data.
Figure 1. Widely implemented ML and DL algorithms in previous studies.
Figure 2. ML and DL algorithms that achieved the best results in each study.
Table 4. Summary of the previous studies that used clinical data.

4. Challenges and Opportunities

4.1. Identifying the Disease

Predicting preeclampsia early and precisely is critical because it influences treatment response and prevents long-term complications in pregnancies. Preeclampsia has a complex disease presentation due to the lack of clear symptoms visible to the patient, and the signs are mostly silent. Choosing the right vital signs for early-stage diagnosis is a significant and challenging step in identifying the disease. Therefore, it is essential to consult doctors to select the right features, test the results, and ensure that the diagnosis is accurate, as inefficient health information systems can contribute to diagnostic errors. For instance, Tahir et al. [12] collected several references of papers which discussed preeclampsia as an initial step for choosing the features, but it resulted in having a diverse set of features. The researchers overcame this challenge by consulting obstetrics and gynecology specialists who assisted in choosing the right attributes and reasoning their choices.

4.2. Patients’ Data Security and Privacy

Patients’ data privacy may limit data sharing, hindering the development of precise ML models in medicine and limiting its progress. As a result, a solution to regulate data sharing in a way that does not impede progress, induce biases against underrepresented populations, or violate any patient privacy laws or regulations must be utilized [38]. A common solution is federated learning, which is a method of training AI models without allowing anyone else to access or touch the data, allowing you to unlock information to feed new AI applications [39]. Implementing robust security measures and protocols is of utmost importance to guarantee the confidentiality and dependability of healthcare systems that gather patient information. Given the sensitive nature of such data, it is imperative to safeguard it against any breaches that could lead to data exposure in further studies [40].

4.3. Reliability of the Models

In a clinical setting, the reliability of ML models includes performance metrics such as sensitivity, accuracy, specificity, precision, and AUC. Sensitivity is crucial to medical studies, since having high sensitivity is necessary to miss as few positive cases as possible [41]. Furthermore, ensuring ethical fairness and relevance for clinical translation is essential to ensure a reliable model and gain the trust and acceptance of health workers and patients. Thus, explainable artificial intelligence (XAI) practices can be used to interpret the black-box models and assist healthcare workers with translation [42].

4.4. Issues Related to the Datasets

Having the proper dataset size is essential to prevent overfitting and underfitting data [43]. Missing values in the dataset can also negatively impact the ML algorithms’ performance and the models’ accuracy [44]. Moreover, in [45], the study included a low prevalence of the disease in the population studied, resulting in an imbalanced dataset. This can cause bias in the results and make it harder for the algorithm to accurately classify the data. Several studies had a considerable drawback of having limited data sizes [46,47,48]. Furthermore, a number of these studies only had access to data from a single center [29,49], which could have resulted in a biased outcome. To improve the accuracy and trustworthiness of diagnostic models, it is advisable to utilize larger and multicenter datasets.
Furthermore, in [31], a small dataset was used in developing the models, which makes it difficult to interpret the results and questions the reliability of the models. To prevent such scenarios, the synthetic minority oversampling technique can be used to avoid an imbalanced dataset [45], and understanding the data’s distribution and handling the missing data based on the distribution minimizes this problem [44].

4.5. Model Interpretation

The ML models are black-box models, which means that the models are so complicated that humans cannot easily read them. The study in [22] highlighted the challenge of implementing ML models in clinical practice due to their complexity, arising from the potential inclusion of a large number of biomarkers, which can make the model more difficult to interpret and use in day-to-day medical decision making. In healthcare, where many decisions are truly life and death, a lack of interpretability in prediction models might weaken trust in those models [46]. XAI is considered a solution to improve the comprehension and interpretation of the predictions made by an ML model [47].

4.6. Human Barriers with AI Adoption in Healthcare

Despite the promising advances in utilizing AI in healthcare, human barriers arise while ensuring quality assurance and accuracy when adopting these technologies. The challenge of assigning liability in this situation is further complicated by the fact that AI technologies are constantly evolving and improving. As such, it is difficult to determine whether the algorithm or the physician should be held liable for any missed findings. Enhancing the human–computer interaction may reduce the cost of this problem, as algorithmic interpretability gives a better understanding in AI’s decisions [48].

4.7. Model Bias

This challenge refers to the risk of ML systems reflecting and amplifying societal biases, leading to unequal accuracy in minority subgroups. This can result in unfair outcomes in areas such as medicine where hospital mortality prediction algorithms may show varying accuracy. To address this issue, it is important to ensure that the data used for training the model are diverse and representative of the target population. Additionally, performance analysis should be conducted by considering subgroups such as age, ethnicity, and location to identify any potential biases in the model. As a result, XAI techniques [47] can be used to make the model’s decision-making process more transparent and explainable, allowing for further examination and adjustment to prevent any biases from being amplified.

5. Conclusions

This review attempted to offer a thorough analysis of the prior achievements made by researchers in the field of preeclampsia prediction. The use of ML algorithms and AI technologies in the medical industry has improved preeclampsia prediction applications. By identifying many ML, DL, and general techniques for preeclampsia prediction, we discovered that the most popular methods were RF, SVM, and LR. In the future, using real datasets, ML and DL algorithms can be utilized to forecast the disease. These datasets can include demographic, clinical, and laboratory information. Moreover, the ensemble method is the creation of a strong collaborative overall model by combining multiple models; this strategy can be utilized to enhance the overall prediction outcomes [49]. Lastly, the evaluation metrics used in the studies to evaluate the model results included AUC, ROC, confusion matrix, accuracy, specificity, recall (also known as sensitivity), F1 score, and precision, and the results were validated using K-fold cross-validation and LOO cross-validation.

Author Contributions

Conceptualization, S.S.A.; methodology, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; software, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; validation, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; formal analysis, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; investigation, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; resources, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; data curation, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; writing—original draft preparation, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A. and H.S.; writing—review and editing, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A., H.S., N.A., I.U.K., D.A.A. and A.A.; visualization, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A., H.S., N.A., I.U.K., D.A.A. and A.A.; supervision, S.S.A.; project administration, S.S.A., N.A., I.U.K., D.A.A. and A.A; funding acquisition, S.S.A., M.A. (Manar Alzahrani), R.A., M.A. (Majd Altukhais), S.A., H.S., N.A., I.U.K., D.A.A. and A.A. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Fabjan-Vodusek, V.; Kumer, K.; Osredkar, J.; Verdenik, I.; Gersak, K.; Premru-Srsen, T. Correlation between Uterine Artery Doppler and the SFlt-1/PlGF Ratio in Different Phenotypes of Placental Dysfunction. Hypertens. Pregnancy 2019, 38, 32–40. [Google Scholar] [CrossRef]
  2. Alrowaili, M.M.; Zakari, N.M.A.; Hamadi, H.Y.; Moawed, S. Management of Gestational Hypertension Disorders in Saudi Arabia by Primary Care Nurses. Saudi Crit. Care J. 2020, 4, 103. [Google Scholar] [CrossRef]
  3. Roberts, J.M.; Gammill, H.S. Preeclampsia. Hypertension 2005, 46, 1243–1249. [Google Scholar] [CrossRef]
  4. Pelícia, S.M.D.C.; Fekete, S.M.W.; Corrente, J.E.; Rugolo, L.M.S.D.S. Impact of Early-Onset Preeclampsia on Feeding Tolerance and Growth of Very Low Birth Weight Infants during Hospitalization. Rev. Paul. Pediatr. 2023, 41, e2021203. [Google Scholar] [CrossRef]
  5. Govender, S.; Naicker, T. The Contribution of Complement Protein C1q in COVID-19 and HIV Infection Comorbid with Preeclampsia: A Review. Int. Arch. Allergy Immunol. 2022, 183, 1114–1126. [Google Scholar] [CrossRef]
  6. Rokotyanskaya, E.A.; Panova, I.A.; Malyshkina, A.I.; Fetisova, I.N.; Fetisov, N.S.; Kharlamova, N.V.; Kuligina, M.V. Technologies for Prediction of Preeclampsia. Sovrem. Tehnol. V Med. 2020, 12, 78–86. [Google Scholar] [CrossRef]
  7. Soomro, T.A.; Zheng, L.; Afifi, A.J.; Ali, A.; Yin, M.; Gao, J. Artificial Intelligence (AI) for Medical Imaging to Combat Coronavirus Disease (COVID-19): A Detailed Review with Direction for Future Research. Artif. Intell. Rev. 2022, 55, 1409–1439. [Google Scholar] [CrossRef]
  8. Hamet, P.; Tremblay, J. Artificial Intelligence in Medicine. Metabolism 2017, 69, S36–S40. [Google Scholar] [CrossRef]
  9. Kumar, N.; Aggarwal, D. LEARNING-Based Focused WEB Crawler. IETE J. Res. 2021, 67, 1–9. [Google Scholar] [CrossRef]
  10. Xue, Y.; Chen, S.; Qin, J.; Liu, Y.; Huang, B.; Chen, H. Application of Deep Learning in Automated Analysis of Molecular Images in Cancer: A Survey. Contrast Media Mol. Imaging 2017, 2017, 9512370. [Google Scholar] [CrossRef]
  11. Bakator, M.; Radosav, D. Deep Learning and Medical Diagnosis: A Review of Literature. Multimodal Technol. Interact. 2018, 2, 47. [Google Scholar] [CrossRef]
  12. Tahir, M.; Badriyah, T.; Syarif, I. Classification Algorithms of Maternal Risk Detection For Preeclampsia With Hypertension During Pregnancy Using Particle Swarm Optimization. EMITTER Int. J. Eng. Technol. 2018, 6, 236–253. [Google Scholar] [CrossRef]
  13. Kodepogu, K.R.; Annam, J.R.; Vipparla, A.; Krishna, B.V.N.V.S.; Kumar, N.; Viswanathan, R.; Gaddala, L.K.; Chandanapalli, S.K. A Novel Deep Convolutional Neural Network for Diagnosis of Skin Disease. Traitement Du Signal 2022, 39, 1873–1877. [Google Scholar] [CrossRef]
  14. Soongsatitanon, A.; Phupong, V. Prediction of Preeclampsia Using First Trimester Placental Protein 13 and Uterine Artery Doppler. J. Matern. Fetal Neonatal Med. 2022, 35, 4412–4417. [Google Scholar] [CrossRef]
  15. Serra, B.; Mendoza, M.; Scazzocchio, E.; Meler, E.; Nolla, M.; Sabrià, E.; Rodríguez, I.; Carreras, E. A New Model for Screening for Early-Onset Preeclampsia. Am. J. Obstet. Gynecol. 2020, 222, e1–e608. [Google Scholar] [CrossRef]
  16. Byonanuwe, S.; Fajardo, Y.; Nápoles, D.; Alvarez, A.; Cèspedes, Y.; Ssebuufu, R. Predicting Risk of Chronic Hypertension in Women with Preeclampsia Based on Placenta Histology. A Prospective Cohort Study in Cuba. 2020. Available online: https://www.researchsquare.com/article/rs-44764/v1 (accessed on 12 December 2022).
  17. Modak, R.; Pal, A.; Pal, A.; Ghosh, M.K. Prediction of Preeclampsia by a Combination of Maternal Spot Urinary Protein-Creatinine Ratio and Uterine Artery Doppler. Int. J. Reprod. Contracept. Obstet. Gynecol. 2020, 9, 635. [Google Scholar] [CrossRef]
  18. Marić, I.; Tsur, A.; Aghaeepour, N.; Montanari, A.; Stevenson, D.K.; Shaw, G.M.; Winn, V.D. Early Prediction of Preeclampsia via Machine Learning. Am. J. Obstet. Gynecol. MFM 2020, 2, 100100. [Google Scholar] [CrossRef]
  19. Jhee, J.H.; Lee, S.; Park, Y.; Lee, S.E.; Kim, Y.A.; Kang, S.-W.; Kwon, J.-Y.; Park, J.T. Prediction Model Development of Late-Onset Preeclampsia Using Machine Learning-Based Methods. PLoS ONE 2019, 14, e0221202. [Google Scholar] [CrossRef]
  20. Marin, I.; Pavaloiu, B.-I.; Marian, C.-V.; Racovita, V.; Goga, N. Early Detection of Preeclampsia Based on a Machine Learning Approach. In Proceedings of the 2019 E-Health and Bioengineering Conference (EHB), Iasi, Romania, 21–23 November 2019; pp. 1–4. [Google Scholar]
  21. Liu, M.; Yang, X.; Chen, G.; Ding, Y.; Shi, M.; Sun, L.; Huang, Z.; Liu, J.; Liu, T.; Yan, R.; 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. [Google Scholar] [CrossRef]
  22. Li, Y.; Shen, X.; Yang, C.; Cao, Z.; Du, R.; Yu, M.; Wang, J.; Wang, M. Novel Electronic Health Records Applied for Prediction of Pre-Eclampsia: Machine-Learning Algorithms. Pregnancy Hypertens. 2021, 26, 102–109. [Google Scholar] [CrossRef]
  23. Carreno, J.F.; Qiu, P. Feature Selection Algorithms for Predicting Preeclampsia: A Comparative Approach. In Proceedings of the 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Seoul, Republic of Korea, 16–19 December 2020; pp. 2626–2631. [Google Scholar]
  24. 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—Goodtechs ’18, Bologna, Italy, 28–30 November 2018; ACM Press: New York, NY, USA, 2018; pp. 232–237. [Google Scholar]
  25. Bosschieter, T.M.; Xu, Z.; Lan, H.; Lengerich, B.J.; Nori, H.; Sitcov, K.; Souter, V.; Caruana, R. Using Interpretable Machine Learning to Predict Maternal and Fetal Outcomes. arXiv 2022, arXiv:2207.05322. [Google Scholar]
  26. Schmidt, L.J.; Rieger, O.; Neznansky, M.; Hackelöer, M.; Dröge, L.A.; Henrich, W.; Higgins, D.; Verlohren, S. A Machine-Learning–Based Algorithm Improves Prediction of Preeclampsia-Associated Adverse Outcomes. Am. J. Obstet. Gynecol. 2022, 227, e1–e77. [Google Scholar] [CrossRef]
  27. Sufriyana, H.; Wu, Y.-W.; Su, E.C.-Y. Prediction of Preeclampsia and Intrauterine Growth Restriction: Development of Machine Learning Models on a Prospective Cohort. JMIR Med. Inform. 2020, 8, e15411. [Google Scholar] [CrossRef]
  28. Sufriyana, H.; Wu, Y.-W.; Su, E.C.-Y. Artificial 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. [Google Scholar] [CrossRef]
  29. Zhang, X.; Chen, Y.; Salerno, S.; Li, Y.; Zhou, L.; Zeng, X.; Li, H. Prediction of Severe Preeclampsia in Machine Learning. Med. Nov. Technol. Devices 2022, 15, 100158. [Google Scholar] [CrossRef]
  30. Lin, Y.C.; Mallia, D.; Clark-sevilla, A.O.; Catto, A.; Leshchenko, A.; Haas, D.M.; Raja, A.; Salleb-aouissi, A. Preeclampsia Predictor with Machine Learning: A Comprehensive and Bias-Free Machine Learning Pipeline. medRxiv 2022. [Google Scholar] [CrossRef]
  31. Villalaín, C.; Herraiz, I.; Domínguez-Del Olmo, P.; Angulo, P.; Ayala, J.L.; Galindo, A. Prediction of Delivery Within 7 Days After Diagnosis of Early Onset Preeclampsia Using Machine-Learning Models. Front. Cardiovasc. Med. 2022, 9, 910701. [Google Scholar] [CrossRef]
  32. Tahir, M.; Badriyah, T.; Syarif, I. Neural Networks Algorithm to Inquire Previous Preeclampsia Factors in Women with Chronic Hypertension During Pregnancy in Childbirth Process. In Proceedings of the 2018 International Electronics Symposium on Knowledge Creation and Intelligent Computing (IES-KCIC), Bali, Indonesia, 29–30 October 2018; pp. 51–55. [Google Scholar]
  33. Sakinah, N.; Tahir, M.; Badriyah, T.; Syarif, I. LSTM With Adam Optimization-Powered High Accuracy Preeclampsia Classification. In Proceedings of the 2019 International Electronics Symposium (IES), Surabaya, Indonesia, 27–28 September 2019; pp. 314–319. [Google Scholar]
  34. Manoochehri, Z.; Manoochehri, S.; Soltani, F.; Tapak, L.; Sadeghifar, M. Predicting Preeclampsia and Related Risk Factors Using Data Mining Approaches: A Cross-Sectional Study. Int. J. Reprod. Biomed. 2021, 19, 959–968. [Google Scholar] [CrossRef]
  35. Han, Q.; Zheng, W.; Guo, X.D.; Zhang, D.; Liu, H.F.; Yu, L.; Yan, J.Y. A New Predicting Model of Preeclampsia Based on Peripheral Blood Test Value. Eur. Rev. Med. Pharmacol. Sci. 2020, 24, 7222–7229. [Google Scholar]
  36. Bennett, R.; Mulla, Z.D.; Parikh, P.; Hauspurg, A.; Razzaghi, T. An Imbalance-Aware Deep Neural Network for Early Prediction of Preeclampsia. PLoS ONE 2022, 17, e0266042. [Google Scholar] [CrossRef]
  37. Dugoff, L.; Hobbins, J.C.; Malone, F.D.; Porter, T.F.; Luthy, D.; Comstock, C.H.; Hankins, G.; Berkowitz, R.L.; Merkatz, I.; Craigo, S.D.; et al. First-Trimester Maternal Serum PAPP-A and Free-Beta Subunit Human Chorionic Gonadotropin Concentrations and Nuchal Translucency Are Associated with Obstetric Complications: A Population-Based Screening Study (The FASTER Trial). Am. J. Obstet. Gynecol. 2004, 191, 1446–1451. [Google Scholar] [CrossRef]
  38. Seastedt, K.P.; Schwab, P.; O’Brien, Z.; Wakida, E.; Herrera, K.; Marcelo, P.G.F.; Agha-Mir-Salim, L.; Frigola, X.B.; Ndulue, E.B.; Marcelo, A.; et al. Global Healthcare Fairness: We Should Be Sharing More, Not Less, Data. PLOS Digit. Health 2022, 1, e0000102. [Google Scholar] [CrossRef]
  39. Yang, Q.; Liu, Y.; Cheng, Y.; Kang, Y.; Chen, T.; Yu, H. Federated Learning; Synthesis Lectures on Artificial Intelligence and Machine Learning; Springer: Berlin/Heidelberg, Germany, 2019; Volume 13, pp. 1–207. [Google Scholar] [CrossRef]
  40. Jiang, L.; Wu, Z.; Xu, X.; Zhan, Y.; Jin, X.; Wang, L.; Qiu, Y. Opportunities and Challenges of Artificial Intelligence in the Medical Field: Current Application, Emerging Problems, and Problem-Solving Strategies. J. Int. Med. Res. 2021, 49, 030006052110001. [Google Scholar] [CrossRef]
  41. Hicks, S.A.; Strümke, I.; Thambawita, V.; Hammou, M.; Riegler, M.A.; Halvorsen, P.; Parasa, S. On Evaluation Metrics for Medical Applications of Artificial Intelligence. Sci. Rep. 2022, 12, 5979. [Google Scholar] [CrossRef]
  42. Balagurunathan, Y.; Mitchell, R.; El Naqa, I. Requirements and Reliability of AI in the Medical Context. Phys. Med. 2021, 83, 72–78. [Google Scholar] [CrossRef]
  43. Zhang, H.; Zhang, L.; Jiang, Y. Overfitting and Underfitting Analysis for Deep Learning Based End-to-End Communication Systems. In Proceedings of the 2019 11th International Conference on Wireless Communications and Signal Processing (WCSP), Xi’an, China, 23–25 October 2019; pp. 1–6. [Google Scholar]
  44. Mishra, P.; Biancolillo, A.; Roger, J.M.; Marini, F.; Rutledge, D.N. New Data Preprocessing Trends Based on Ensemble of Multiple Preprocessing Techniques. TrAC Trends Anal. Chem. 2020, 132, 116045. [Google Scholar] [CrossRef]
  45. Feng, S.; Keung, J.; Yu, X.; Xiao, Y.; Zhang, M. Investigation on the Stability of SMOTE-Based Oversampling Techniques in Software Defect Prediction. Inf. Softw. Technol. 2021, 139, 106662. [Google Scholar] [CrossRef]
  46. Petch, J.; Di, S.; Nelson, W. Opening the Black Box: The Promise and Limitations of Explainable Machine Learning in Cardiology. Can. J. Cardiol. 2022, 38, 204–213. [Google Scholar] [CrossRef]
  47. Speith, T. A Review of Taxonomies of Explainable Artificial Intelligence (XAI) Methods. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, Seoul, Republic of Korea, 21–24 June 2022; ACM: New York, NY, USA; pp. 2239–2250. [Google Scholar]
  48. Grant, K.; McParland, A.; Mehta, S.; Ackery, A.D. Artificial Intelligence in Emergency Medicine: Surmountable Barriers with Revolutionary Potential. Ann. Emerg. Med. 2020, 75, 721–726. [Google Scholar] [CrossRef]
  49. Atallah, R.; Al-Mousa, A. Heart Disease Detection Using Machine Learning Majority Voting Ensemble Method. In Proceedings of the 2019 2nd International Conference on new Trends in Computing Sciences (ICTCS), Amman, Jordan, 9–11 October 2019; pp. 1–6. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Journal Statistics
Multiple requests from the same IP address are counted as one view.
Download PDF
An Improved Link Prediction Approach for Directed Complex Networks Using Stochastic Block Modeling
Previous
The Art of the Masses: Overviews on the Collective Visual Heritage through Convolutional Neural Networks
Next