Machine learning predictive system to predict the risk of developing pre-eclampsia

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Pre-eclampsia is a significant cause of maternal and fetal morbidity.
• Accurate and early risk assessment is crucial for timely intervention.
• Existing prediction methods may be costly or less accessible.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based predictive model for pre-eclampsia risk.
• To utilize routinely collected clinical data for model development.
• To identify key clinical features predictive of pre-eclampsia.

Main Methods:

• Retrospective analysis of 2444 pregnant women's clinical data (2015-2019).
• Development of five ML models: logistic regression, random forest, light gradient boosting machine, extreme gradient boosting (XGBoost), and multilayer perceptron.
• Application of Synthetic Minority Oversampling Technique (SMOTE) and SHapley Additive exPlanations (SHAP) for feature importance.

Main Results:

• XGBoost demonstrated superior performance with an area under the receiver operating characteristic curve (AUC) of 0.921.
• Key predictors identified by SHAP analysis include diastolic blood pressure, systolic blood pressure, and urine glucose.
• The model achieved high accuracy, sensitivity, and specificity.

Conclusions:

• Machine learning, specifically XGBoost, effectively predicts pre-eclampsia risk using standard clinical data.
• This ML approach provides a cost-effective alternative to expensive diagnostic tests.
• The developed model facilitates real-time risk assessment and supports early clinical intervention.