JMIR Publications

ABSTRACT

Background:

Fetal growth restriction (FGR) is a common complication of preeclampsia (PE). FGR in patients with PE increases the risk of neonatal perinatal mortality and morbidity. However, early prediction methods for FGR are biased or clinically unexplainable, which is difficult to apply to clinical practice, leading to a relative delay in intervention and a lack of effective treatments.

Objective:

The study aims to develop an auxiliary diagnostic model based on machine learning (ML) to predict the occurrence of FGR in PE patients.

Methods:

This study utilized a retrospective case-control approach to analyze 38 features, including the basic medical history and peripheral blood laboratory test results of pregnant patients with PE, either or not complicated by FGR. ML models were constructed to evaluate the predictive value of maternal parameter changes on the PE combined with FGR. Multiple algorithms were tested, including logistic regression, light gradient boosting (LGB), random forest (RF), extreme gradient boosting (XGB), multilayer perceptron (MLP), naive Bayes (NB) and support vector machine (SVM). The model performance was identified by the area under the receiver operating characteristic curve (AUC). The SHapley Additive exPlanation method (SHAP) was adopted to rank the feature importance and explain the final model for clinical application.

Results:

The random forest (RF) model performed best in discriminative ability among the 7 ML models. After reducing features according to importance rank, an explainable final RF model was established with 9 features, including Urinary protein quantification(UPQ), Gestational week of delivery (GWD), Umbilical artery Systolic-to-Diastolic Ratio (S/D), Amniotic fluid Index (AFI), Triglyceride, D-dimer, Weight, Height and Maximum systolic pressure (MSP). The model could accurately predict FGR for 513 PE patients in the training and testing dataset (AUC = 0.83±0.03, n=149 FGR and 364 without FGR) using 5-fold cross-validation, which was closely validated for 103 PE patients in an external dataset(AUC = 0.82±0.048, n=45 FGR and 58 without FGR). On the whole, UPQ, Umbilical artery S/D and GWD exhibited the highest contributions to the model performance (c=0.45,0.34 and 0.33) based on SHAP analysis. For specific individual patient, SHAP analysis can reveal the protective and risk factors to develop RGR for interpretating the model’s clinical significance. Finally, the model has been translated into a convenient webpage tool to facilitate its utility in clinical settings.

Conclusions:

We successfully developed a model that accurately predicts FGR development in patients with PE. The SHAP method captures highly relevant risk factors for model interpretation, alleviating concerns about the “black box” problem of indirect interpretation of ML techniques.