Machine-learning-based prediction of pre-eclampsia using first-trimester maternal characteristics and biomarkers

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Maternal-Fetal Medicine

Background:

• Pre-eclampsia (PE) is a significant cause of maternal and fetal morbidity.
• Accurate early prediction of PE is crucial for timely intervention.
• Current prediction methods often rely on limited factors.

Purpose of the Study:

• To assess the accuracy of AI and machine learning in predicting PE risk.
• To evaluate the contribution of maternal demographics, medical history, and biomarkers.
• To determine the impact of including or excluding race in prediction models.

Main Methods:

• Utilized a large dataset from 11-13 weeks gestation for training and validation.
• Employed an artificial neural network incorporating maternal factors and biomarkers (UtA-PI, MAP, PlGF, PAPP-A).
• Assessed prediction accuracy using Area Under the Curve (AUC) and detection rates at various false-positive rates (FPR).

Main Results:

• AI models combining maternal factors and biomarkers achieved high prediction accuracy for preterm PE (AUC 0.909).
• Excluding race significantly reduced prediction accuracy, particularly for non-white groups.
• High MAP, UtA-PI, and low PlGF were key predictors; aspirin use was recommended for high-risk cases.

Conclusions:

• Machine learning offers a robust, population-independent method for PE screening.
• Race is a critical factor for accurate PE risk prediction, especially when using maternal characteristics alone.
• AI-driven prediction models enhance early detection and management of pre-eclampsia.