基于人工神经网络的巨大儿预测的研究  

Prediction of fet al macrosomia with BP neural network

作　　者：申爱方[1] 张玲[1] 由卫芝[1] 李少丛[1] 焦盈盈[1] 

机构地区：[1]山东省聊城市人民医院脑科医院,252000

出　　处：《中国实用医药》2011年第26期33-34,共2页China Practical Medicine

摘　　要：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 investigate the value of BP neural network in predicting fetal macrosomia. Methods The 390 cases of full-term singleton pregnancy without complication were divided into two groups,100 samples for training and 290 samples for testing. Used the height, weight, uterine fundal height and abdominal circumference of pregnant women as well as the bipariet al diameter（BPD）,femur length（FL）and amniotic fluid index（AFI）.If the output value above 0.5,it indicate fetal macrosomia. Results The BP method had the predicting accuracy of 75%,better than other method. Conclusion The fetal weight estimation using artificial neural networks shows a potential prospect of application. A suitable output value is critical to build up an effective network.

关 键 词：巨大儿 预测 人工神经网络 

分 类 号：R714.5[医药卫生—妇产科学] TP183[医药卫生—临床医学]