Construction of Predictive Model for Aided Diagnosis of Fetal Growth Restriction Based on Ultrasonic Blood Flow Parameters and Screening of Machine Learning Algorithm
LIU Bingbing
ZHU Haohui
WANG Ruili
GAO Yuan
ZHANG Xiaolin
CHEN Juan
CHEN Xin
Abstract：Objective To construct a predictive model of fetal growth restriction(FGR)assisted diagnosis based on ultrasonic blood flow parameters using machine learning algorithm,and compare the predictive efficiency of different models.Methods Data of 88 pregnant women with FGR who were admitted to Henan Provincial People's Hospital from January 2021 to December 2023 were included in this study(FGR group),and 88 non-FGR pregnant women who gave birth in Henan Provincial People's Hospital were used as controls(non-FGR group).In this study,univariate and multi-factor logistic regression analysis was used to screen the risk factors affecting FGR,and R software was used to construct support vector machine prediction models,random forest prediction models and logistic regression prediction models.The receiver operating characteristic curves of three prediction models were drawn,and the effectiveness of different models in predicting FGR was compared to determine the optimal prediction tool.Results Gestational weeks,middle cerebral artery pulsatility index(MCAPI),umbilical artery pulsatility index(UAPI),ductus venosus pulsatility index(DVPI)and cerebroplacental ratio(CPR)were significantly different between FGR group and non-FGR group(P＜0.05).There were no significant differences in age,maternal type,body mass index and newborn sex(P＞0.05).Logistic regression analysis showed that gestational week,DVPI and CPR were the risk factors for FGR(P＜0.05).Based on the three risk factors affecting FGR,logistic regression prediction model,support vector machine prediction model and random forest prediction model were constructed respectively.The results showed that the random forest prediction model had the best curve performance,and its area under the curve was 0.986(95％CI:0.975-0.997).Conclusion Gestational age,DVPI and CPR are all risk factors for FGR,and the random forest prediction model is the most effective in predicting FGR.
Keywords：fetal growth restrictionDoppler ultrasoundblood flow parametersmachine learning algorithmpredictive value
Publication Date：2024-11-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：4( 3868-3871 )