A New Birth Weight Prediction Model Based On Ultrasonography At Term

Part ? Correlation analysis between fetal fractional thigh volume and fractional upper arm volume and neonatal birth weight at termObjectivesTo investigate the correlation between neonatal birth weight(BW)and fractional thigh volume(TVol),BW and fractional upper arm volume(AVol)obtained by three dimensional(3D)ultrasonography at term.MethodsAntepartum fetal TVol and AVol data sets from normal singleton pregnant women in the third trimester(at 37?41+6 weeks of gestation)were collected by 3D ultrasonography before delivery.TVol and AVol values were calculated through off-line analysis of volume data sets above by 4D View software.Intra-and interclass coefficients(ICC)were calculated from 40 random cases of TVol and AVol by the same operator or different operators.The reproducibility and consistency of fetal TVol and AVol measurements were examined by Bland and Altman difference method.Pearson correlation analysis and linear regression analysis were used to find out the relationship between BW and TVol,BW and AVol.P<0.05 was considered statistically significant.ResultsThe ICC(intra-group and inter-group)calculated from the 40 cases of TVol and AVol values obtained by the same operator and different operators were all higher than 0.90 and the average percentage differences in the Bland-Altman difference method were less than 50%.The results of Pearson correlation analysis between BW and TVol,BW and AVol were rrvoi = 0.87,rAVol = 0.85(P<0.05),respectively.The equation of antepartum TVol and BW through linear regression analysis was BWTVol = 40.77 ×TVol + 1199.92(RTVol2 = 0.77).And the equation of antepartum AVol and BW before labor was BWAVol = 80.59 × AVol + 1431.27(RAVol2 = 0.76).ConclusionThere is a positive correlation between full-term fetal TVol and BW,AVol and BW,both of which are important indicators of neonatal birth weight.Part II A new birth weight prediction model based on two dimensional and three dimensional ultrasonography through PCA at termObjectivesTo investigate the correlations between fetal growth measurement parameters obtained by two dimensional(2D)and three dimensional(3D)ultrasonography.Based on the correlations above,to establish a new neonatal birth weight estimation(EFW)model through principal component analysis(PCA)at term.Methods2D and 3D ultrasonography was prospectively performed in normal singleton pregnant women in the third trimester(37?41+6 weeks of gestation)within 7 days before delivery.Fetal measurement parameters were recorded.Intra-and interclass coefficients(ICC)were calculated from 40 random cases of 2D and 3D ultrasonographic measurement data sets by the same operator or different operators.The reproducibility and consistency of fetal 2D and 3D ultrasonographic measurements were examined by Bland and Altman difference method.All of the participants were divided into the EFW model development group(n=600)and the validation group(n=597)randomly.PC A was used to analyze the multicollinearity among all the nine ultrasonographic parameters in the EFW model development group.And a new PCA model was developed combing application of stepwise linear regression analysis.The new model's accuracy in predicting fetal birth weight was confirmed by the EFW model validation group through calculation of mean percentage error(MPE)and predictive rate of ABW±1%,5%,and 10%,in comparisons with previously published common used bith weight prediction formulae.P<0.05 was considered statistically significant.ResultsThe ICC(intra-group and inter-group)calculated from the 40 cases of 2D and 3D ultrasonographic parameters obtained by the same operator and different operators were all higher than 0.75 and the average percentage differences in the Bland-Altman difference method were not significant(P>0.05).In the EFW model development group,the PCA results showed that the Kaiser-Meyer-Olkin measure of sampling adequacy was found to be 0.86 for the multicollinearity among the ultrasonographic parameters and the Chi-Square value of Bartlett's test of sphericity was 3931.306(P<0.05).Two principal components(PC1 and PC2)were extracted from 2D and 3D ultrasonographic parameters and the cumulative contribution rate reached 72 372%.A new PCA model was established according to the relationalship.In the EFW model development group,the new PCA model has a MPE of 3.661%and a standard deviation(SD)of 3.161%.Compared with previously published EFW formulae,at least a 47.558%decrease in the MPE and a 57.421%decrease in the SD of the new PCA model were noted.The results were similar to those in the EFW model validation group and the predictive rate of ABW±1%,5%,and 10%in the new model were 15.58%,73.20%,and 100.00%,respectively.ConclusionsIn this study,PCA analysis was used to confirm and eliminate the multicollinearity between 2D and 3D ultrasonographic parameters.We obtain comprehensive variables representing fetal soft tissue and osseous parts,respectively,and a more accurate birth weight prediction model.