The Research On Methods Of Estimation Of Fetal Weight Based On Ultrasound And Analysis Of Its Related Factors

Objectives To explore the best model for estimation of fetal weight of full-term fetus based on ultrasonography.To find out the critical reference values of different biological parameters and maternal parameters when macrosomia was born.To analyze risk factors of macrosomia.Methods A total of 407 term pregnant women with singleton between October 2015 and January 2016 in the North China University of Science and Technology Affiliated Hospital were included in the present study.An ultrasound examination was performed 5 days before delivery.They were divided into non-macrosomia group(FW<4000g,n = 337)and macrosomia group(FW?4000g,n=70).Clinical data(height,weight,weight gain during pregnancy,gestational age,uterine height,abdominal circumference etc.)and ultrasound data(biparietal diameter,occipitofrontal diameter,head circumference,liver length,transverse cerebellar diameter,transverse diameter of abdomen,anteroposterior diameter of abdomen,abdominal circumference,femur length and mid-thigh soft tissue thickness)were collected.Database was established by Excel 2013,SPSS 20 statistical software was uesd for statistical analysis.The normal distribution of datas were measurement by((?)±s),measurement data of skew distribution with a median(four point interval)representation.Correlation analysis of the two groups of data using Pearson correlation analysis.New equations for estimation of fetal weight were established by multiple linear regression analysis.Comparison of measurement data using ANOVA,t-test,rank-sum test,intragroup pairwise compared by LSD test.Logistic regression analysis was used to analyze the risk factors of macrosomia.Test level alpha P<0.05.Results 1 Comparison the accuracy of existing 19 regression equations for estimation of fetal weight:(1)In the 5 clinical parameter equations,when the non-macrosomia group and overall group using Zhuo Jingru method,macrosomia group using Luo Laimin method,the absolute errors and relative errors is less than the other 4 equations(P<0.05),then the accuracy of estimation of fetal weight is higher than that of the other 4 equations.(2)In the14 ultrasonic parameter equations,when the non-macrosomia group and overall group using Hadlock FP(BPD,HC,AC,FL),macrosomia group using Merz E(BPD,AC),the absolute errors and relative errors is less than the other 13 equations(P<0.05),then the accuracy of estimation of fetal weight is higher than that of the other 13 equations.2 New equations for estimation of fetal weight were established:(1)Correlation analysis of the parameters and fetal weight: AC(r=0.806,P<0.05)was closely related to fetal weight in the correlation between maternal and fetal parameters and fetal weight.(2)Three equations are established: clinical parameters: New Equation 1;ultrasonic parameters: New Equation2;joint parameter: New Equation 3;(3)In the 3 new equations,when all group using New Equation 3,the absolute errors and relative errors is less than the other 2 new equations(P<0.05),then the accuracy of estimation of fetal weight is higher than that of the other 2equations.3 In all 22 equations,when all group using the New Equation 3,the absolute errors and relative errors is less than the other 21 equations(P<0.05),then the accuracy forestimation of fetal weight is the highest.4 BP artificial neural network models for estimation of fetal weight:(1)The absolute error and relative error of BP artificial neural network model with the joint parameters decreased with the increase of the number of training samples in a certain range(P<0.05).The results show that the number of training samples of BP neural network can improve the accuracy of predicting fetal weight.(2)The absolute error and relative error of the BP artificial neural network model with ultrasonic parameters or joint parameters for estimation of fetal weight in different groups is less than that of the equations(P<0.05),then the accuracy for estimation of fetal weight were higher than that of the regression equation.(3)When estimation of fetal weight,the absolute error and relative error of the BP neural network model of joint parameters in the nonmacrosomia group and the overall group is less than the clinical parameters and ultrasonic parameters of BP artificial neural network model(P<0.05).The absolute error and relative error of the BP neural network model with joint parameter and ultrasonic parameters in the macrosomia group is less than the clinical parameters(P<0.05),but there is no difference between the two.The results show that the joint parameter BP artificial neural network model has the highest accuracyfor estimation of fetal weight.5 The analysis of ROC curve for prediction of macrosomia with different parameters: When the value of fundal height is35.5cm,the sensitivity and specificity to predict macrosomia was 73.7%,82.2%;when the value of TCD is 5.34 cm,the sensitivity and specificity to predict macrosomia was 85.4%,92.3%.The results showed that uterine height,TCD have high sensitivity and specificity to predict macrosomia.6 Multivariate analysis of macrosomia: high levels of blood glucose(OR=1.440,95%CI 1.063~1.950,P<0.05),high levels of triglycerides(OR=1.212,95%CI 1.068~1.375,P<0.05),high body mass index(OR=1.208,95%CI 1.113~1.742,P<0.05),high weight gain index during pregnancy(OR=1.113,95%CI 1.013~1.223,P<0.05)were the risk factors of macrosomia,high levels of LDLC(OR=0.625,95%CI0.431~0.908,P<0.05)is a protective factor for appearance of macrosomia.Conclusions 1 When using the existing equations to predict fetal weight,we should choose the appropriate equation according to the different weight range of the fetus.2 In different range of fetal weight,the newly established joint parameter equation New Equation 3 can accurately predict fetal birth weight.3 The optimal model for estimation of fetal weight was the the joint parameter BP artificial neural network model with high training samples.4 The best clinical and ultrasound parameters of prediction macrosomia:uterine height,TCD.5 Risk factors for macrosomia: pregnant women had high body mass index,high weight gain index during pregnancy,high levels of triglycerides,high levels of blood glucose,but high levels of LDLC were protective factors.