| 张雅娜,和平,王晶晶,夏太慧,吴青青.基于超声标准化描述联合多元分析建立预测子宫内膜良恶性病变模型[J].中国医学影像技术,2021,37(3):350~354 |
| 基于超声标准化描述联合多元分析建立预测子宫内膜良恶性病变模型 |
| Establishment of predictive model of benign and malignant endometrial lesions based on ultrasonic standardized description combined with multivariate analysis |
| 投稿时间：2020-12-15 修订日期：2021-02-27 |
| DOI：10.13929/j.issn.1003-3289.2021.03.008 |
| 中文关键词: 子宫内膜 标准化 多元分析 超声检查 |
| 英文关键词:endometrium standardized multivariate analysis ultrasonography |
| 基金项目: |
|
| 摘要点击次数: 2482 |
| 全文下载次数: 847 |
| 中文摘要: |
| 目的 筛选有助于鉴别绝经后且子宫内膜厚度（ET）≥ 5 mm女性子宫内膜良恶性病变的超声及临床指标，建立子宫内膜癌个体患病风险的Logistic回归模型。方法 回顾性分析261例绝经后且ET ≥ 5 mm子宫内膜病变患者标准化超声声像图特征及临床信息，根据病变性质分为良性组（n=157）及恶性组（n=104），组内按2：1比例分为建模亚组和验证亚组，以单因素分析筛选鉴别子宫良恶性病变具有价值的指标，利用Logistic回归建立模型预测良恶性病变。结果 良性组与恶性组患者年龄、绝经年限、体质量指数（BMI）、高血压史、绝经后阴道出血症状、ET、内膜回声不均且不伴小囊、内膜线不规则或显示不清、内膜-肌层交界中断及CDFI显示存在多支血流信号情况差异均有统计学意义（P均<0.05）。最终采用BMI>24 kg/m2、阴道出血症状、ET>8 mm、内膜-肌层交界中断及多支血流信号5个指标建立回归分析模型，其受试者工作特征（ROC）曲线下面积（AUC）为0.905；以0.33为最佳截断值，模型预测良恶性病变准确率、敏感度、特异度、阳性预测值、阴性预测值及约登指数分别为82.18%、92.80%、75.20%、71.11%、94.05%及68.00%。结论 应用临床及超声特征建立了预测绝经后女性子宫内膜癌发病风险模型，可为个性化诊疗提供帮助。 |
| 英文摘要: |
| Objective To screen ultrasound and clinical indicators able to identify benign and malignant endometrial lesions in postmenopausal women with endometrial thickness (ET) ≥ 5 mm, and to establish a logistic regression model to predict the individual risk of endometrial cancer. Methods Standardized ultrasonic features and clinical information of 261 postmenopausal patients with endometrial lesions and ET ≥ 5 mm were retrospectively analyzed. The patients were divided into benign group (n=157) and malignant group (n=104) according to pathological results, and then further divided into modeling subgroup and validation subgroup on the ratio of 2:1. Univariate analysis was used to screen the valuable indicators for differentiating benign and malignant uterine lesions, and Logistic regression was used to establish the prediction model. Results There were statistical differences of age, years since menopause, body mass index (BMI), hypertension, postmenopausal bleeding, ET, non-uniform endometrial echogenicity without cystic area, irregular or not defined endometrial midline, interrupted endometrial-myometrial junction and multiple vessels on CDFI between 2 groups (all P<0.05). The final Logistic regression model was established based on BMI>24 kg/m2, postmenopausal bleeding, ET>8 mm, interrupted endometrial-myometrial junction and multiple vessels on CDFI. The area under the receiver operating characteristics (ROC) curve (AUC) was 0.905. Taken 0.33 as the best cutoff value, the accuracy, sensitivity, specificity, positive predictive value, negative predictive value and Youden index for detection of endometrial malignant lesions was 82.18%, 92.80%, 75.20%, 71.11%, 94.05% and 68.00%, respectively. Conclusion The regression model established using clinical and ultrasonic characteristics for predicting the risk of endometrial cancer in postmenopausal women was helpful to personalized diagnosis and treatment. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
