多模态超声成像在乳腺影像报告和数据系统4类乳腺病变中的应用:恶性预测模型
本研究纳入454例经组织学检查证实的BI-RADS 4类乳腺病变患者,利用常规超声、应变弹性成像和超声造影建立并验证恶性预测模型。模型在队列1、2、3的AUC分别为0.967、0.948和0.920,选择性降级4a类病变可使特异度从0.0%提高至80.0%(队列2)和77.6%(队列3),敏感度无显著下降。
要点
- 年龄≥40岁、超声上纵横比大于1、超声造影早期高增强和增强面积增大是乳腺恶性的独立危险因素。
- 多模态超声预测模型对BI-RADS 4类病变的恶性判别具有良好区分度,验证队列AUC为0.948和0.920。
- 利用模型选择性降级4a类病变可提高特异度、减少不必要的活检,但存在敏感度损失。
关键影像表现/征象
- 常规超声上纵横比大于1(taller than wide shape)是乳腺恶性的独立危险因素(p<0.05)。
- 超声造影早期高增强是乳腺恶性的独立危险因素(p<0.05)。
- 超声造影增强面积增大是乳腺恶性的独立危险因素(p<0.05)。
诊断标准与关键数值
- 预测模型logistic回归方程:p = 1/1+Exp∑[-5.066 + 3.125 x(年龄≥40岁)+ 1.943 x(纵横比大于1)+ 1.479 x(早期高增强)+ 4.167 x(增强面积增大)]。
- 预测模型AUC在队列1为0.967,队列2为0.948,队列3为0.920。
- 利用预测模型选择性降级4a类病变后,重新评定的BI-RADS在队列2的AUC为0.880(95%CI 0.794-0.965),队列3为0.870(95%CI 0.801-0.939)。
- 队列2特异度从0.0%(0/35)提高至80.0%(28/35),敏感度从100.0%降至95.9%(p=0.153)。
- 队列3特异度从0.0%(0/58)提高至77.6%(45/58),敏感度从100.0%降至96.4%(p=0.081)。
临床应用与阅片要点
- 作者认为,将多模态超声加入超声BI-RADS可提高BI-RADS 4类病变的特异度,减少不必要的活检。
- 作者指出,利用预测模型选择性降级4a类病变时存在敏感度损失。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
The purpose of this study was to develop, validate and test a prediction model for discriminating malignant from benign breast lesions using conventional ultrasound (US), US elastography of strain elastography and contrast-enhanced ultrasound (CEUS). The study included 454 patients with breast imaging-reporting and data system (BI-RADS) category 4 breast lesions identified on histologic examinations. Firstly, 228 breast lesions (cohort 1) were analyzed by logistic regression analysis to identify the risk factors, and a breast malignancy prediction model was created. Secondly, the prediction model was validated in cohort 2 (84 patients) and tested in cohort 3 (142 patients) by using analysis of the area under the receiver operating characteristic curve (AUC). Univariate regression indicated that age ≥40 y, taller than wide shape on US, early hyperenhancement on CEUS and enlargement of enhancement area on CEUS were independent risk factors for breast malignancy (all p < 0.05). The logistic regression equation was established as follows: p = 1/1+Exp∑[-5.066 + 3.125 x (if age ≥40 y) + 1.943 x (if taller than wide shape) + 1.479 x (if early hyperenhancement) + 4.167 x (if enlargement of enhancement area). The prediction model showed good discrimination performance with an AUC of 0.967 in cohort 1, 0.948 in cohort 2 and 0.920 in cohort 3. By using the prediction model to selectively downgrade category 4a lesions, the re-rated BI-RADS yield an AUC of 0.880 (95% confidence interval [CI], 0.794-0.965) in cohort 2 and 0.870 (95% CI, 0.801-0.939) in cohort 3. The specificity increased from 0.0% (0/35) to 80.0% (28/35) without loss of sensitivity (from 100.0% to 95.9%, p = 0.153) in cohort 2. Similarly, the specificity increased from 0.0% (0/58) to 77.6% (45/58) without loss of sensitivity (from 100.0% to 96.4%, p = 0.081) in cohort 3. Multimodal US showed good diagnostic performance in predicting breast malignancy of BI-RADS category 4 lesions. Although the loss of sensitivity was existing, the addition of multimodal US to US BI-RADS could improve the specificity in BI-RADS category 4 lesions, which reduced unnecessary biopsies.
引用
- Li XL, Lu F, Zhu AQ, et al. Multimodal Ultrasound Imaging in Breast Imaging-Reporting and Data System 4 Breast Lesions: A Prediction Model for Malignancy. Ultrasound Med Biol. 2020. PMID: 32896449. doi:10.1016/j.ultrasmedbio.2020.08.003
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。