Prediction of Preeclampsia Using Machine Learning: A Systematic Review

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Preeclampsia significantly contributes to maternal and perinatal morbidity and mortality worldwide.
• Early prediction of preeclampsia is crucial for timely intervention, such as aspirin prophylaxis.
• Machine learning (ML) is emerging as a powerful tool for disease prediction and prognosis.

Purpose of the Study:

• To review methodologies, predictors, and performance of ML models for preeclampsia prediction.
• To emphasize comparative advantages, challenges, and clinical applicability of ML in preeclampsia.
• To explore the potential of artificial intelligence and deep learning in managing preeclampsia.

Main Methods:

• Systematic literature search of PubMed, Cochrane, and Scopus (last 10 years).
• Keywords included "preeclampsia", "risk factors", "machine learning", "artificial intelligence", and "deep learning".
• 11 eligible studies were included after screening 325 records, assessing risk of bias.

Main Results:

• A wide range of predictors were used, including clinicodemographic data, lab results, Doppler ultrasound, genotypic data, and fundal images.
• Over ten different ML models were employed across studies from various countries.
• Models like XGBoost, random forest, and neural networks showed high predictive accuracy (AUC 0.76-0.97).

Conclusions:

• ML models demonstrate significant potential for accurate preeclampsia prediction.
• Interpretability of 'black box' models is a critical ethical consideration for clinical adoption.
• Future research should focus on diverse population validation and balancing performance with interpretability for improved maternal outcomes.