目的探讨BP神经网络对巨大儿预测的意义。方法将病例分为训练组100例,验证组290例,输入层有孕妇身高、体重、宫高、腹围;B超下胎儿双顶径、股骨长径、羊水指数;构建神经网络,输出层0.5以上为巨大儿。结果符合率为75%,优于临床常用方法。结论 BP神经网络预测巨大儿有良好的应用前景,选取良好的输出层可提高预测的准确性。 Objective To explore the significance of BP neural network in predicting macrosomia. Methods The cases were divided into training group (n = 100), validation group (n = 290), pregnant women with height, weight, uterine height and abdominal circumference at the input layer. B-fetus fetus with biparietal diameter, long femur diameter and amniotic fluid index 0.5 or more for the giant children. Results The coincidence rate was 75%, better than clinical common methods. Conclusion BP neural network prediction of giant children has a good prospect, select a good output layer can improve the accuracy of prediction.