基于影像组学构建乳腺癌前哨淋巴结转移预测模型的研究  

Radiomics based on DCE-MRI for the preoperative prediction of SLN metastasis in breast cancer

作　　者：刘梅婕 毛宁[2] 马恒 史英红[2] 董建军[2] 杨平[2] 张堃 车凯莉 段绍峰 张学喜 谢海柱[2] LIU Meijie;MAO Ning;MA Heng;SHI Yinghong;DONG Jianjun;YANG Ping;ZHANG Kun;CHE Kaili;DUAN Shaofeng;ZHANG Xuexi;XIE Haizhu(Department of Radiology,Yantai Yuhuangding Hospital,Yantai,264000,China)

机构地区：[1]滨州医学院临床医学院医学影像学系,山东烟台264000 [2]青岛大学附属烟台毓璜顶医院影像科,山东烟台264000 [3]GE Healthcare China,上海200000

出　　处：《中国中西医结合影像学杂志》2020年第3期227-231,共5页Chinese Imaging Journal of Integrated Traditional and Western Medicine

基　　金：山东省自然科学基金项目(ZR2017PH043)。

摘　　要：目的:探讨基于MRI动态增强扫描(DCE-MRI)的影像组学在预测乳腺癌前哨淋巴结(SLN)转移中的价值。方法:回顾性收集经病理证实并行DCE-MRI检查的浸润性乳腺癌164例(训练组124例,验证组40例)。在DCE-MRI图像上提取影像组学特征,并计算DCE参数,采用Lasso-Logistic回归模型对影像组学特征进行筛选。分别建立单纯影像组学模型、单纯DCE参数模型及联合模型。采用ROC的AUC评价不同模型的鉴别预测效能,并对模型的ROC曲线行DeLong检验;在验证队列中评估其预测效能。结果:共提取396个影像组学特征,经筛选得到28个特征,联合DCE参数分别建模。对于术前预测SLN转移的效能,在训练组中单纯影像组学模型AUC的95%CI为0.81(0.72,0.89),单纯DCE参数模型AUC的95%CI为0.77(0.68,0.86),联合预测模型AUC的95%CI为0.80(0.72,0.89);在验证组中单纯影像组学模型AUC的95%CI为0.74(0.59,0.89),单纯DCE参数模型AUC的95%CI为0.74(0.59,0.90),联合预测模型AUC的95%CI为0.76(0.61,0.91),Delong检验显示差异无统计学意义(P>0.05),联合模型的效能可能稍高。结论:基于DCE-MRI图像提取影像组学特征及DCE参数建立预测模型,作为一种无创性预测乳腺癌SLN转移的工具,有良好的应用前景。Objective:To establish a radiomic model based on dynamic contrast enhanced MRI(DCE-MRI)for predicting sentinel lymph node(SLN)metastasis in patients with breast cancer.Methods:A total of 164 patients(124 cases in the training cohort and 40 cases in the validation cohort)were retrospectively enrolled,all confirmed by pathology and underwent conventional DCE-MRI before surgery.Radiomic features and hemodynamics characteristics were derived from DCE-MRI data.Lasso-Logistic regression models were established for data dimension reduction,feature selection.Three prediction models based on hemodynamic characteristics alone,radiomics signature alone,and their combination were developed.AUC of ROC was used to evaluate the predictive effectiveness of the model and verified by the validation cohort data.Results:28 radiomic features were selected to construct the radiomics signature.In the training cohort,radiomics signature,the hemodynamic characteristics model,and the combined model yielded AUC(95%CI)values of 0.81(0.72,0.89),0.77(0.68,0.86),and 0.80(0.72,0.89),respectively.In the validation cohort,radiomics signature,the hemodynamic characteristics model,and the combined model yielded AUC(95%CI)values of 0.74(0.59,0.89),0.74(0.59,0.90),and 0.76(0.61,0.91),respectively.Combined model added more net benefit than either feature alone.Conclusion:The radiomics model incorporating radiomics signature and hemodynamic characteristics can be conveniently used to facilitate preoperative individualized prediction of SLN metastasis in patients with breast cancer.

关 键 词：乳腺癌 前哨淋巴结 动态增强磁共振 影像组学 DCE参数 

分 类 号：R737.9[医药卫生—肿瘤] R730.44[医药卫生—临床医学] R445.2