| 摘要: |
| 目的 建立基于临床资料、剪切波弹性成像参数和超声影像组学的列线图模型，探讨其鉴别BI-RADS 4类乳腺病变良恶性的效能。方法 回顾性收集2017年12月至2023年6月3家医院共403例BI-RADS 4类乳腺病变患者的临床资料、剪切波弹性成像及病理检查结果，以2017年12月至2019年6月南京鼓楼医院和2019年6月至2019年12月安徽医科大学第一附属医院共283个乳腺病灶为训练集，2022年4月至2023年6月北京世纪坛医院120个乳腺病灶为验证集，按病理结果，将训练集和验证集分为良性组和恶性组。通过提取病灶灰阶超声影像组学特征计算影像组学评分（Rad-score）。采用单因素及多因素Logistic回归分析鉴别BI-RADS 4类乳腺病变良恶性的影响因素，构建预测模型并绘制列线图，采用受试者工作特征（ROC）曲线、校准曲线及临床决策曲线评估该模型的效能。结果 经过特征提取及筛选，最终纳入13个影像组学特征用于计算Rad-score，验证集良、恶性组Rad-score分别为[-1.07 (-1.64, -0.37)分、0.07（-0.3，0.56）分]，二者比较差异有统计学意义（Z=514，P<0.001）。多因素Logistic回归分析显示年龄（OR值：1.107，P<0.001）和最大剪切波速度（SWVmax）（OR值：3.919，P<0.001）及Rad-score（OR值：4.18，P<0.001）是预测乳腺恶性病变的独立影响因素。基于以上3个因素构建的列线图模型在训练集中及验证集中鉴别BI-RADS 4类乳腺病变良恶性的ROC曲线下面积均高于SWVmax和Rad-score（均P＜0.001），且拟合度均良好（均P＞0.05）；在验证集中使用列线图模型预测BI-RADS 4类病变能获得更高的临床收益，将非必要穿刺活检率降低了61.16%。结论 基于患者年龄、SWVmax及Rad-score构建的列线图模型能有效预测BI-RADS 4类乳腺病变良恶性，降低非必要穿刺活检率，有一定的临床价值。 |
| 关键词: 超声检查 剪切波弹性成像 影像组学 列线图 BI-RADS 4类 良恶性 |
| DOI： |
| 投稿时间：2023-08-24修订日期：2023-12-10 |
| 基金项目:南京市科技发展计划项目（编号：201803027） |
|
| Improving US-BI-RADS 4 lesions assessment：A nomogram combining radiomics，clinical and shear wave elastography |
| Zhou Weijun,yangmin,liuyong,xuping,wulanying,wangying,Xiao rong |
| (Department of ultrasound,Beijing Shijitan Hospital,Capital Medical University) |
| Abstract: |
| Objective To develop a nomogram model based on radiomics，clinical and shear wave elastography（SWE）, and to explore the efficiency of the model in differentiating benign and malignant lesions in BI-RADS 4 category lesions. Methods From December 2017 to June 2023, 403 patients with BI-RADS 4 category lesions from 3 hospitals were retrospectively reported. The clinical characteristics，SWE parameters and pathological results were collected. A total of 283 breast lesions in Nanjing Drum Tower Hospital and the First Affiliated Hospital of Anhui Medical University from December 2017 to December 2019 were divided into training set. A total of 120 breast lesions in Beijing Shijitan Hospital from April 2022 to June 2023 were divided into the validation set. According to pathological results, the training set and the validation set were divided into benign and malignant group. The radiomics features of B-mode ultrasound of the lesions were extracted and the radiomics score（Rad-score） was calculated. Univariate and multivariate logistic regression analysis were used to determine the independent predictors of breast cancer and construct a nomogram. The receiver operating characteristic curve，calibration curve and clinical decision curve were used to evaluate the diagnostic efficacy of the nomogram. Results After feature extraction and screening, 13 radiomics features were finally included in the calculation of Rad-score. The Rad-score of the benign and malignant groups were [-1.07 (-1.64, -0.37) points and 0.07 (-0.3, 0.56), respectively], and the difference between them was statistically significant (Z=514,P＜0.001). Multivariate Logistic regression analysis showed that age (OR: 1.107, P＜0.001), maximum shear wave velocity (SWVmax) (OR: 3.919, P＜0.001) and Rad-score (OR: 4.18, P＜0.001) were independent factors in predicting malignant breast lesions. The nomogram model was conducted with age，SWVmax and Rad-score. The area under ROC curve of the nomogram model was higher than that of SWVmax and Rad-score (both P＜0.001) in the training set and validation set, and the fit was good (both P＞0.05). In the validation set, the nomogram model could achieve higher clinical benefits to predict BI-RADS 4 category lesions and could reduce the non-necessary biopsy rate of BI-RADS 4 category lesions by 61.2%. Conclusion The nomogram model based on age, SWVmax and Rad-score can effectively predict the benign and malignant lesions of BI-RADS 4 category lesions, and reduce the rate of unnecessary biopsy. |
| Key words: Ultrasound examination shear wave elastography radiomics nomogram BI-RADS 4 categories Benign and malignant |
