| 摘要: |
| 目的 使用logistic回归模型及CHAID决策树模型分析胎儿染色体异常的影响因素,并比较两种模型的优劣。方法 回顾性分析超声软指标（ultrasound soft marker,USM）阳性并具有羊水穿刺结果的642例单胎孕妇资料，以胎儿染色体结果为因变量，USM为自变量建立logistic回归模型及决策树模型筛选影响胎儿染色体结果的因素，绘制ROC曲线比较两种模型的效果。结果 单因素logistic回归模型显示NT增厚、鼻骨缺失、侧脑室增宽为胎儿染色体异常的危险因素；多因素logistic回归分析筛选NT增厚（OR=7.511，P<0.05）、鼻骨缺失（OR=4.819，P<0.05）、侧脑室增宽（OR=4.789，P<0.05）3个因素用于回归模型的拟合；CHAID决策树模型显示NT增厚、鼻骨缺失是胎儿染色体异常的影响因素；logistic回归模型ROC曲线下面积大于CHAID决策树模型（0.712vs 0.675，Z=2.267，P<0.05）。结论 logistic回归模型及决策树模型对胎儿染色体结果有一定的预测价值，且logistic回归模型优于决策树模型。
关键词：logistic回归模型；CHAID决策树模型；胎儿染色体异常；超声软指标 |
| 关键词: logistic回归模型 CHAID决策树模型 胎儿染色体异常 超声软指标 |
| DOI： |
| 投稿时间：2022-04-14修订日期：2022-05-21 |
| 基金项目:]陆军特色医学中心人才创新能力培养项目(2019CXJXC014),重庆市影像医学与核医学临床医学研究中心(项目编号:CSTC2015YFPT-gcjsyjzx0175)。 |
|
| Prediction of fetal chromosomal abnormalities based on logistic regression model and CHAID decision tree model |
| yangrong,LuoXiaoyong,LiTao |
| () |
| Abstract: |
| Objective To analyze the influencing factors of fetal chromosomal abnormalities by logistic regression model and Chi-squared Automatic Interaction Detector (CHAID) decision tree model, and compare the advantages and disadvantages of the two statistical methods. Methods The data of 642 singleton pregnant women with positive ultrasound soft marker and amniocentesis results were retrospectively analyzed,with fetal chromosomal abnormalities as the dependent variable and ultrasound soft marker as independent variables, logistic regression model and decision tree model had been established to screen the factors affecting fetal chromosomal abnormalities, and receiver operating characteristic (ROC)curve was drawn in order to see the differences of the effects of the two models. Results The univariate logistic regression model showed that the nuchal translucency thickening, absent nasal bone, and ventriculomegaly were the risk factors for fetal chromosomal abnormalities. Three factors were used to fit the regression model, these factors including nuchal translucency thickening（OR=7.511，P<0.05）, absent nasal bone (OR=4.819，P<0.05) and ventriculomegaly (OR=4.789，P<0.05) were selected by the multivariate logistic regression model. The decision tree model showed that the nuchal translucency thickening and absent nasal bone were the influencing factors of fetal chromosomal abnormalities. The ROC curve area of the logistic regression model was larger while that of the decision tree model show a smaller area (0.712vs 0.675，Z=2.267，P<0.05). Conclusion Logistic regression model and decision tree model have certain value in analyzing the relationship between USM and fetal chromosomal abnormalities, and the logistic regression model is better than the decision tree model.
Key words
logistic regression model; CHAID decision tree model; fetal chromosomal abnormality; ultrasound soft marker |
| Key words: logistic regression model CHAID decision tree model fetal chromosomal abnormality ultrasound soft marker |
