标题：基于自动乳腺全容积成像构建的列线图模型对HER-2阳性乳腺癌的预测价值

      作者：陈泉，胡萍，陈良，何燕    昆山市中医医院超声科，江苏 昆山 215300

      卷次： 2024年35卷12期

      【摘要】 目的 探讨自动乳腺全容积成像(ABVS)技术对人类表皮生长因子受体 2 (HER-2)阳性乳腺癌的预测价值，构建列线图模型并验证。方法 选取2018年8月至2023年10月在昆山市中医医院术前行ABVS检查并最终病理证实为乳腺癌的女性患者216例，将216例乳腺癌患者按7∶3比例随机分成训练集151例和验证集65例。根据免疫组化HER-2表达水平将训练集分为HER-2阴性组(n=94)和HER-2阳性组(n=57)，验证集分为HER-2阴性组(n=45)和HER-2阳性组(n=20)。比较训练集两组乳腺癌病灶的超声特征，采用多因素 Logistic回归筛选出HER-2阳性乳腺癌的独立预测因素，构建预测HER-2阳性乳腺癌的列线图模型。采用受试者工作特征(ROC)曲线、校准曲线、决策曲线分析(DCA)评估模型在训练集和验证集中的预测效能。结果 训练集两组乳腺癌在边缘、后方回声、微钙化、汇聚征之间比较差异均有统计学意义(P<0.05)；经多因素Logistic回归分析结果显示，边缘成角、微钙化、后方回声无衰减、冠状面无汇聚征是HER-2阳性乳腺癌的独立预测因素(P<0.05)；根据上述独立预测因素构建列线图模型，训练集的曲线下面积(AUC)为0.843 (95%CI：0.778~0.908)，特异度为72.3%，敏感度为87.7%，验证集的AUC为0.789 (95%CI：0.667~0.911)，特异度为73.3%，敏感度为80.0%；训练集及验证集的校准曲线和DCA曲线显示该列线图模型具有较好的校准度和临床应用价值。结论 基于ABVS构建的列线图模型对HER-2阳性乳腺癌具有良好的预测性能，可以辅助病理诊断并协助临床制定乳腺癌患者的治疗方案。
      【关键词】 乳腺癌；自动乳腺全容积成像；人类表皮生长因子受体2；列线图
      【中图分类号】 R737.9 【文献标识码】 A 【文章编号】 1003—6350（2024）12—1771—05

Predictive value of nomogram model based on automated breast volume scanner for HER-2-positive breastcancer.
CHEN Quan, HU Ping, CHEN Liang, HE Yan. Department of Ultrasound, Kunshan Traditional Chinese MedicineHospital, Kunshan 215300, Jiangsu, CHINA
【Abstract】 Objective To explore the predictive value of automated breast volume scanner (ABVS) for humanepidermal growth factor receptor 2 (HER-2)-positive breast cancer, and to construct a nomogram model and verify it.Methods A total of 216 female patients who received ABVS examination in Kunshan Traditional Chinese MedicineHospital and were pathologically confirmed as breast cancer from August 2018 to October 2023 were selected. The pa-tients were randomly divided into 151 training sets and 65 validation sets according to the ratio of 7:3. According to theexpression level of HER-2 in immunohistochemistry, the training set was divided into HER-2-negative group (n=94)and HER-2-positive group (n=57), and the validation set was divided into HER-2-negative group (n=45) andHER-2-positive group (n=20). The ultrasonic characteristics of breast cancer lesions were compared between the twogroups of training sets, the independent predictors of HER-2 positive breast cancer were screened by multivariate logis-tic regression, and a nomogram model for predicting HER-2 positive breast cancer was constructed. The predictive per-formance of the model in the training and validation sets was evaluated using receiver operating characteristic (ROC)curve, calibration curve, and decision curve analysis (DCA). Results There were statistically significant differences inmargin, posterior echo, microcalcification, and retraction phenomenon between HER-2-negative group andHER-2-positive group of the training set (P<0.05). Multivariate logistic regression analysis showed that angular margin,microcalcification, no attenuation of posterior echo, and no retraction phenomenon on coronal plane were independentpredictors of HER-2-positive breast cancer (P<0.05). Based on the independent predictive factors mentioned above, anomgram model was constructed. The area under the curve (AUC) of the training set was 0.843 (95% CI: 0.778-0.908),with a specificity of 72.3% and a sensitivity of 87.7%. The AUC of the validation set was 0.789 (95% CI: 0.667-0.911),with a specificity of 73.3% and a sensitivity of 80.0%. The calibration curves and DCA curves of the training and valida-tion sets showed that the nomgram model has good calibration and clinical application value. Conclusion The nomo-gram model based on ABVS has good prediction performance for HER-2-positive breast cancer, which can assist inpathological diagnosis and clinical development of treatment plans for breast cancer patients.
      【Key words】 Breast cancer; Automated breast volume scanner; Human epidermal growth factor receptor 2; Nom-gram