术前MRI影像组学列线图对乳腺癌腋窝淋巴结转移的预测价值

Predictive value of preoperative MRI-based nomogram for axillary lymph node metastasis in breast cancer

目的建立基于术前乳腺MRI的影像组学列线图,探讨MRI影像组学模型对乳腺癌腋窝淋巴结转移(axillary lymph node metastasis,ALNM)的预测价值。材料与方法回顾性分析宁夏医科大学总医院2016年8月至2020年12月经病理确诊的169例女性乳腺癌患者的MRI及临床病理资料(118例为训练集,51例为验证集)。在动态对比增强MRI第3期图像中,对所有患者的乳腺肿瘤原发灶进行感兴趣区(volume of interest,VOI)勾画并提取影像组学特征。采用Mann-Whitney U检验、LASSO回归进行影像组学特征筛选并建立影像组学标签,对所选特征按LASSO回归中相应系数加权后求和来计算影像组学评分。同时通过Logistic回归筛选临床危险因素建立临床预测模型,并构建基于临床危险因素和影像组学标签的联合预测模型。使用受试者工作特征(receiver operating characteristic,ROC)曲线和校准曲线评估模型性能,使用Delong检验评价不同模型间ROC曲线下面积(area under curve,AUC)的差异,使用决策曲线分析(decision curve analysis,DCA)评估预测模型的临床价值。结果从每个VOI中提取200个影像组学特征,其中10个影像组学特征与乳腺癌发生ALNM相关。在训练集及验证集中,影像组学标签的AUC值分别为0.86及0.74;临床预测模型的AUC值分别为0.83、0.78;联合预测模型的AUC值分别为0.86及0.80。DCA显示联合模型的临床价值高于影像组学标签及临床模型,使用列线图能够可视化该联合预测模型。结论基于术前MRI影像组学标签和临床危险因素所构建的联合模型可用于乳腺癌ALNM的预测,为术前评估乳腺癌患者发生ALNM的风险提供了一种新的方法。

Objective:To establish a preoperative breast MRI-based radiomics nomogram and to explore the predictive value of the MRI-based radiomics model for axillary lymph node metastasis(ALNM) in breast cancer.Materials and Methods:Between August2016 and December 2020,the MRI and clinicopathological data of 169 female patients(training set,n=118;validation set,n=51)identified as breast cancer by pathological examination were retrospectively collected and analyzed in the General Hospital of Ningxia Medical University.In the third phase of dynamic contrast-enhanced MRI,a volume of interest(VOI) of the primary breast tumor in each patient was delineated,and then the radiomics features of the VOI were extracted.The Mann-Whitney U test and LASSO regression were used to select the radiomics features and establish radiomics signature.The selected features were weighted by their corresponding coefficients of LASSO regression and then summed to calculate the radiomics score.The Logistic regression was used to select the clinical risk factors and establish a clinical predictive model.In addition,a combined predictive model including the clinical risk factors and radiomics signatures were constructed.The receiver operating characteristic(ROC) curve and the calibration curve were used to evaluate the performances of the models.The Delong test was used to compare the differences of the area under curve(AUC) values among different predictive models.The decision curve analysis(DCA) was conducted to assess the clinical use of these predictive models.Results:Among 200 radiomics features extracted from each VOI,10 of them were associated with the presence of ALNM in breast cancer patients.In the training set and validation set,the radiomics signature had an AUC value of 0.86 and 0.74,respectively;the clinical predictive model had an AUC value of 0.83 and 0.78,respectively;the combined predictive model had an AUC value of 0.86 and0.80,respectively.DCA showed the clinical use of these three predictive models.The combined predictive model could be visualized by a nomogram.Conclusions:A combined model incorporating preoperative MRI-based radiomics signatures and clinical risk factors can be used to predict the presence of ALNM in breast cancer.It provides a new method to preoperative assess the risk of the presence of ALNM in patients with breast cancer.