超声预测足月胎儿出生体重方法的探究及其相关因素分析

发布时间：2018-03-23 23:08

  本文选题：胎儿体重　切入点：回归方程　出处：《华北理工大学》2017年硕士论文

【摘要】：目的探讨基于超声检查预测足月胎儿出生体重的最佳模型;分析寻找出现巨大儿时临床参数、超声参数的临界参考值;分析寻找出现巨大儿的危险因素。方法选取2015年10月至2017年01月在华北理工大学附属医院产科住院分娩的单胎孕足月孕妇407例,均在产前0-5天进行胎儿超声检查。将407例胎儿分为:A组(非巨大儿组FW4000g),337例;B组(巨大儿组FW≥4000g),70例。超声测量参数包括:双顶径-BPD、枕额径-OFD、头围-HC、小脑横径-TCD、肝脏长度-LL、腹横径、腹前后径、腹围-AC、股骨长度-FL及股骨中段皮下软组织厚度-FSTT等。收集的临床参数包括:孕妇身高、体重、孕期增重,孕周、宫高、腹围,生化指标等。采用Excel 2013建立数据库,SPSS 20.0进行统计学分析,正态分布的资料以((?)±s)表示,偏态资料以中位数(四分位数间距)表示。各参数与胎儿体重的相关性分析采用Pearson相关性分析;建立新的预测胎儿体重回归方程采用多重线性回归分析法;计量资料间的比较采用单因素方差分析、t检验、秩和检验,组内两两之间比较采用LSD检验,计数资料间的比较采用卡方检验;各参数预测巨大儿临界参考值的确定采用ROC曲线分析;分析巨大儿的危险因素采用Logistic回归分析;以P0.05差异有统计学意义。结果1已有的19种回归方程预测胎儿体重准确性的比较:(1)在5种临床参数方程中,非巨大儿组及整体组采用卓晶如法、巨大儿组采用罗来敏法预测胎儿体重时绝对误差、相对误差均小于其余4个方程(P0.05),表明其预测胎儿体重准确性高于其余4个方程;(2)在14种临床参数方程中,非巨大儿组、整体组采用Hadlock FP(BPD、HC、AC、FL)、巨大儿组采用Merz E(BPD、AC)预测胎儿体重时绝对误差、相对误差均小于其余13个方程(P0.05),表明其预测胎儿体重准确性高于其余13种方程;2建立新的预测胎儿体重回归方程:(1)各参数与胎儿体重的相关性分析:AC(r=0.806,P0.05)与胎儿体重的相关程度最密切;(2)新建立三个回归方程:(1)临床参数方程New Equation 1;(2)超声参数方程:New Equation 2;(3)联合参数方程New Equation 3。(3)在新建立的3种方程中,不同组别采用New Equation 3预测胎儿体重时绝对误差、相对误差均小于其余2个方程(P0.05),表明其预测胎儿体重的准确性高于其余2个新方程;3在所有22种方程中(现有19种方程及新建立的3种方程),不同组别采用New Equation 3预测胎儿体重时绝对误差、相对误差均小于其余21个方程(P0.05),表明New Equation 3为预测胎儿体重准确性最高的回归方程。4 BP人工神经网络模型预测胎儿体重:(1)在训练样本数一定范围内,超声参数、联合参数的BP人工神经网络模型预测胎儿体重绝对误差、相对误差,随着训练样本数的增加而降低(P0.05),表明提高超声参数、联合参数的BP人工神经网络模型训练样本数可以提高预测胎儿体重准确性。(2)在各组别中,超声参数、联合参数的BP人工神经网络模型预测胎儿体重的绝对误差、相对误差均小于回归方程法(P0.05),表明超声参数、联合参数的BP人工神经网络模型预测胎儿体重的准确性高于回归方程法。(3)在非巨大儿组、整体组中,联合参数的BP人工神经网络模型预测胎儿体重的绝对误差、相对误差均小于临床参数、超声参数的BP人工神经网络模型(P0.05);在巨大儿组中,联合参数、超声参数的BP人工神经网络模型预测胎儿体重的绝对误差、相对误差小于临床参数的BP人工神经网络模型(P0.05),但二者之间差异无统计学意义(P0.05),表明在各组中,联合参数BP人工神经网络模型预测胎儿体重准确性最高。5各参数预测巨大儿的ROC曲线分析:当宫高的取值为35.5cm,预测巨大儿的灵敏度、特异度为73.7%、82.2%;当TCD的取值为5.34cm,预测巨大儿的灵敏度、特异度为85.4%、92.3%,表明宫高、TCD对预测巨大儿具有较高的灵敏度及特异度。6巨大儿相关因素分析:高水平血糖(OR=1.440,95%CI 1.063~1.950,P0.05)、高水平甘油三酯(OR=1.212,95%CI 1.068~1.375,P0.05)、高孕妇体重指数(OR=1.208,95%CI 1.113~1.742,P0.05)、高孕期增重指数(OR=1.113,95%CI 1.013~1.223,P0.05)是出现巨大儿的危险因素,高水平LDLC(OR=0.625,95%CI 0.431~0.908,P0.05)是出现巨大儿的保护因素。结论1应用已有的方程预测胎儿体重时,应根据胎儿体重范围选择合适方程。2在各个胎儿体重范围内,New Equation 3能准确预测胎儿出生体重。3预测胎儿体重的最佳模型为高训练样本数的联合参数BP人工神经网络模型。4预测巨大儿发生最佳的临床及超声指标:宫高、TCD。5出现巨大儿的危险因素:高水平血糖、高水平甘油三酯、孕妇高体重指数、孕期高增重指数;而高水平LDLC是保护因素。
[Abstract]:Objective to investigate the ultrasound examination in predicting the best model of fetal birth weight based on the analysis of clinical parameters of childhood; looking for great critical reference value of ultrasonic parameters; analysis for risk factors of macrosomia. Methods from October 2015 to 2017 01 months in North China Polytechnic University hospital obstetrics hospital aboutsingletonnulliparousvertex full-term pregnant women in 407 cases. All fetal ultrasonography in prenatal 0-5 days. 407 fetuses were divided into two groups: group A (non macrosomia group FW4000g, 337 cases); group B (macrosomia group FW = 4000g), including 70 cases. Ultrasound measurement parameters: biparietal diameter -BPD, occipitofrontal diameter -OFD, head -HC, transverse cerebellar diameter -TCD -LL, the length of liver, abdominal diameter, abdominal diameter, abdominal circumference and femur length of -AC, -FL and femoral subcutaneous soft tissue thickness of -FSTT. Clinical parameters were collected: maternal height, body weight, weight gain during pregnancy, pregnancy, uterine height, abdominal circumference, and biochemical indexes by Excel. 2013 to establish a database, SPSS 20 for statistical analysis, the normal distribution of the data in ((?) + s) said that the skewness data to the median (four percentile interval). Correlation analysis of the parameters and the fetal weight by Pearson correlation analysis; the establishment of a new prediction of fetal weight regression equation by multiple linear regression analysis method; the measurement data were analyzed by using the single factor variance analysis, t test, rank sum test between the 22 groups were compared using LSD test, count data were compared by the chi square test; the parameter prediction of macrosomia to determine the critical reference values by ROC curve analysis; analysis of risk factors of macrosomia using Logistic regression analysis on P0.05; there was a significant difference between the results of the 19 regression equation 1. The prediction of the comparison of the accuracy of fetal weight: (1) in the 5 clinical parameters in the equation, the non macrosomia group and the whole group by Zhuo Jingru 娉，

本文编号：1655610