基于MRI影像组学列线图预测乳腺癌腋窝淋巴结转移  

Prediction of axillary lymph node metastasis in breast cancer based on radiomics nomogram of MRI

作　　者：夏旭东 段成洲 李铭[1] 王亚龙 周小山 王功夏 王海彬[1] 崔振华[1] XIA Xudong;DUAN Chengzhou;LI Ming;WANG Yalong;ZHOU Xiaoshan;WANG Gongxia;WANG Haibin;CUI Zhenhua(Department of Radiology,Anyang Tumor Hospital,Anyang 455000,China)

机构地区：[1]安阳市肿瘤医院影像科,安阳455000

出　　处：《磁共振成像》2022年第1期118-122,共5页Chinese Journal of Magnetic Resonance Imaging

基　　金：安阳市重点研发及推广专项项目(编号:20313)。

摘　　要：目的本研究旨在建立和验证基于MRI的影像组学列线图来预测乳腺癌较小体积的腋窝淋巴结(axillary lymph node,ALN)转移。材料与方法回顾性分析2018年1月至2021年4月238例经病理证实的乳腺癌患者,基于动态对比增强磁共振成像(dynamic contrast enhanced magnetic resonance imaging,DCE-MRI)及T2脂肪抑制序列提取ALN纹理特征,采用分层抽样的方式按照7∶3比例分为训练组(n=168)和测试组(n=70),线性回归和LASSO回归用于特征筛选。结合影像组学和MRI影像学特征中的独立因素建立多因素Logistic回归模型,以列线图形式表现。采用受试者工作特征(receiver operating characteristic,ROC)曲线来评价模型的性能。使用Hosmer-Lemeshow检验并绘制校准曲线来评价模型的拟合优度。采用决策曲线分析(decision curve analysis,DCA)评价模型的临床应用价值。结果单因素和多因素分析显示,影像组学标签评分(radiomics score,Rad-score)、短长轴比及ADC值为鉴别淋巴结转移的独立影响因素;Rad-score是最为重要的影响因素(0R=1.413,P<0.001),训练组和测试组ROC曲线下面积(area under the curve,AUC)分别为0.867、0.887;列线图由MRI影像学和影像组学特征组成,该模型显示出良好的校准和区分能力,AUC在训练集中为0.972(95%CI:0.950~0.994),在测试集中为0.938(95%CI:0.882~0.993)。决策曲线分析表明具有临床使用价值。结论基于MRI影像组学列线图可用于乳腺癌ALN转移的术前预测。Objective:To establish and verify a radiomics nomogram based on MRI for predicting small axillary lymph node(ALN)metastasis in breast cancer.Materials and Methods:A retrospective analysis of 238 patients with breast cancer confirmed by pathological from January 2018 to April 2021.ALN texture features were extracted based on dynamic contrast enhanced magnetic resonance imaging(DCE-MRI)and T2 fat suppression sequence,and stratified sampling was used to divide the group into training(n=168)and testing(n=70)groups according to ratio of 7∶3,linear regression and the least absolute shrinkage and selection operator(LASSO)algorithm were used to select the feature.Based on the regression coefficients of the screened features.Multi-factor Logistic regression models combining independent factors from radiomic signature and MR imaging characteristics were developed and presented in the form of nomogram.Receiver operating characteristic(ROC)curve was used to evaluate the performance of the model.Hosmer-Lemeshow test was used and calibration curve was plotted to evaluate the goodness of fit of the model.Decision curve analysis(DCA)was used to evaluate the clinical application value of the model.Results:Univariate and multivariate analysis showed that Rad-score,short-to-long axis ratio and ADC value were independent factors in identifying lymph node metastases.Rad-score was the most important factor(OR=1.413,P<0.001)with areas under the ROC curve(AUC)of 0.867 and 0.887 for the training and testing groups,respectively.The model showed good calibration and discrimination with AUC of 0.972(95%CI:0.950—0.994)in the training set and 0.938(95%CI:0.882—0.993)in the validation set.DCA findings indicated that the nomogram model was clinically useful.Conclusions:The MRI-based radiomics nomogram model could be used to preoperatively predict the ALN metastasis of breast cancer.

关 键 词：乳腺肿瘤 腋窝淋巴结 磁共振成像 影像组学 

分 类 号：R445.2[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学]