论文部分内容阅读
利用神经网络技术,采用易于测量和获得的孕妇参数(宫高、腹围、身高和体重)预测足月妊娠胎儿体重。经152例训练样本训练后,该组预测胎儿体重符合率为86.84%。利用经训练的网络预测140例胎儿体重,符合率为85%。预测胎儿体重与出生时新生儿体重相对误差≤5%和≤1%者分别占预测总数的58.87%的94.28%,利用网络输入对输出贡献分析可知:宫高、腹围、身高和体重4个参数对胎儿体重影响系数分别为67%、13%、16%和4%。其中宫高影响最大。由此可见,该法与现在的采用孕妇参数预测胎儿体重方法中均以宫高为主要参数是一致的。
Full-term fetal weights were predicted using neural network techniques using maternal parameters that are easily measurable and obtained (uterine height, abdominal circumference, height, and weight). After 152 training samples training, the group predicts fetal weight coincidence rate of 86.84%. 140 cases of fetal weight were predicted using a trained network with a coincidence rate of 85%. Predict fetal weight and newborn birth weight at birth ≤ 5% and ≤ 1%, respectively, accounted for 58.87% of the total number of 94.28%, the use of network input output contribution analysis shows that: the uterus, abdominal circumference, height and weight 4 The coefficients of influence on fetal weight were 67%, 13%, 16% and 4% respectively. Among them, Gong Gong has the most influence. Thus, this method and the current method of using fetal parameters to predict fetal weight are high as the main parameter of the palace is consistent.