Fetal weight estimation by automated three-dimensional limb volume model in late third trimester compared to two-dimensional model: a cross-sectional prospective observational study

Background Accurate estimation of fetal weight is important for prenatal care and for detection of fetal growth abnormalities. Prediction of fetal weight entails the indirect measurement of fetal biometry by ultrasound that is then introduced into formulae to calculate the estimated fetal weight. The aim of our study was to evaluate the accuracy of fetal weight estimation of Chinese fetuses in the third trimester using an automated three-dimensional (3D) fractional limb volume model, and to compare this model with the traditional two-dimensional (2D) model. Methods Prospective 2D and 3D ultrasonography were performed among women with singleton pregnancies 7 days before delivery to obtain 2D data, including fetal biparietal diameter, abdominal circumference and femur length, as well as 3D data, including the fractional arm volume (AVol) and fractional thigh volume (TVol). The fetal weight was estimated using the 2D model and the 3D fractional limb volume model respectively. Percentage error was defined as (estimated fetal weight - actual birth weight) divided by actual birth weight and multiplied by 100. Systematic errors (accuracy) were evaluated as the mean percentage error (MPE). Random errors (precision) were calculated as +/- 1 SD of percentage error. The intraclass correlation coefficient (ICC) was used to analyze the inter-observer reliability of the 3D ultrasound measurements of fractional limb volume. Results Ultrasound examination was performed on 56 fetuses at 39.6 +/- 1.4 weeks' gestation. The average birth weight of the newborns was 3393 +/- 530 g. The average fetal weight estimated by the 2D model was 3478 +/- 467 g, and the MPE was 3.2 +/- 8.9. The average fetal weights estimated by AVol and TVol of the 3D model were 3268 +/- 467 g and 3250 +/- 485 g, respectively, and the MPEs were - 3.3 +/- 6.6 and - 3.9 +/- 6.1, respectively. For the 3D TVol model, the proportion of fetuses with estimated error <= 5% was significantly higher than that of the 2D model (55.4% vs. 33.9%, p < 0.05). For fetuses with a birth weight < 3500 g, the accuracy of the AVol and TVol models were better than the 2D model (- 0.8 vs. 7.0 and - 2.8 vs. 7.0, both p < 0.05). Moreover, for these fetuses, the proportions of estimated error <= 5% of the AVol and TVol models were 58.1 and 64.5%, respectively, significantly higher than that of the 2D model (19.4%) (both p < 0.05). The inter-observer reliability of measuring fetal AVol and TVol were high, with the ICCs of 0.921 and 0.963, respectively. Conclusion In this cohort, the automated 3D fractional limb volume model improves the accuracy of weight estimation in most third-trimester fetuses. Prediction accuracy of the 3D model for neonatal BW, particularly < 3500 g was higher than that of the traditional 2D model.