| 杨文惠,杨丽,张月红,吴云,丁莉.超声E-cervix弹性成像参数与传统指标联合模型预测阴道试产结局[J].中国介入影像与治疗学,2026,23(7):415-421 |
| 超声E-cervix弹性成像参数与传统指标联合模型预测阴道试产结局 |
| Combined model based on ultrasound E-cervix elastography parameters and traditional indicators for predicting outcome of vaginal trial of labor |
| 投稿时间：2026-04-16 修订日期：2026-06-26 |
| DOI：10.13929/j.issn.1672-8475.2026.07.007 |
| 中文关键词: 孕妇 弹性成像技术 机器学习 剖宫产术 |
| 英文关键词:pregnant women elasticity imaging techniques machine learning cesarean section |
| 基金项目: |
|
| 摘要点击次数: 39 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的 观察超声E-cervix弹性成像参数与传统指标联合模型预测阴道试产结局的价值。方法 前瞻性纳入170名拟接受阴道试产孕妇，按7∶3比例随机划分训练集（n=118）与验证集（n=52），并根据后续是否转为剖宫产分为顺产亚组（训练集n=83、验证集n=40）与剖宫产亚组（训练集n=35、验证集n=12）；比较亚组间临床及超声资料，筛选阴道试产结局的独立影响因素。分别采用逻辑回归（LR）、支持向量机（SVM）、随机森林（RF）及极限梯度提升（XGBoost）构建联合模型，观察其预测效能。结果 训练集2亚组间分娩史、孕周、宫颈弹性对比指数（ECI）、硬度比（HR）、宫颈内口应变率（IOS）、宫颈外口应变率（EOS）及宫颈长度差异均有统计学意义（P均＜0.05）；分娩史、ECI、HR及宫颈长度均为阴道试产结局的独立影响因素（P均＜0.05）。LR、SVM、RF及XGBoost联合模型预测训练集阴道试产结局的曲线下面积（AUC）分别为0.918、0.913、0.945及0.928，在验证集分别为0.892、0.915、0.927及0.903；均以RF联合模型的AUC最大（P均＜0.05）。结论 基于分娩史、ECI、HR及宫颈长度的RF联合模型可用于预测阴道试产结局。 |
| 英文摘要: |
| Objective To observe the value of combined model based on ultrasound E-cervix elastography parameters and traditional indicators for predicting outcome of vaginal trial of labor. Methods Totally 170 pregnant women who would undergo vaginal trial of labor were prospectively enrolled and randomly allocated to training set （n=118） and validation set （n=52） at a ratio of 7∶3， also into vaginal delivery （83 in training set and 40 in validation set） and cesarean delivery subgroups （35 in training set and 12 in validation set） according to subsequent conversion to cesarean delivery or not. Clinical and ultrasound data were compared between subgroups， and the independent influencing factors of the outcome of vaginal trial of labor were screened. Then combined models were constructed with logistic regression （LR）， support vector machine （SVM）， random forest （RF） and extreme gradient boosting （XGBoost） ， respectively， and their predictive efficacy were evaluated. Results In training set， significant differences of the proportion of delivery history， gestational age， elasticity contrast index （ECI）， hardness ratio （HR）， internal ostium strain （IOS）， external ostium strain （EOS） and cervical length were found between 2 subgroups （all P＜0.05）， and delivery history， ECI， HR and cervical length were all independent influencing factors of the outcome of vaginal trial of labor （all P＜0.05）. The area under the curve （AUC） of LR， SVM， RF and XGBoost combined models for predicting the outcome of vaginal trial of labor was 0.918， 0.913， 0.945 and 0.928 in training set， respectively， while in validation set was 0.892， 0.915， 0.927 and 0.903， respectively. AUC of combined RF model was the highest in both sets （all P＜0.05）. Conclusion Combined RF model based on delivery history， ECI， HR and cervical length could be used to predict the outcome of vaginal trial of labor. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
