表观扩散系数MRI影像组学预测乳腺浸润性导管癌Luminal分型的初步研究  

A preliminary study on the prediction of Luminal typing of breast invasive ductal carcinoma based on apparent diffusion coefficient MRI radiomics

作　　者：李卓琳[1] 赵美琳 丁莹莹[1] 吴建萍[1] 李美地 薛珂 谢瑜[1] LI Zhuolin;ZHAO Meilin;DING Yingying;WU Jianping;LI Meidi;XUE Ke;XIE Yu(Department of Radiology,the Third Affiliated Hospital of Kunming Medical University,Kunming 650018,China)

机构地区：[1]昆明医科大学第三附属医院放射科,云南昆明650018

出　　处：《实用放射学杂志》2021年第6期936-939,948,共5页Journal of Practical Radiology

基　　金：云南省科技厅-昆明医科大学应用基础研究联合专项资金项目(2018FE001(-006))。

摘　　要：目的探讨基于表观扩散系数(ADC)-MRI影像组学模型预测Luminal型与非Luminal型乳腺浸润性导管癌的价值.方法回顾性收集241例经病理证实为乳腺浸润性导管癌患者的临床资料和MRI图像,根据免疫组化和荧光原位杂交技术检测结果分为Luminal A、Luminal B、人类表皮生长因子受体2(HER-2)过表达和基底样型4个分子亚型,Luminal A、B型统称为Luminal型,HER-2过表达和基底样型统称为非Luminal型.将所有病例随机分为训练集(138例)和验证集(103例),在ADC-MRI上勾画病灶的三维感兴趣区(3D-ROI),利用计算机自动化提取影像组学特征;采用Lasso回归模型对特征进行降维、筛选和构建预测模型;通过验证集对预测模型进行验证,运用受试者工作特征(ROC)曲线对预测模型在训练集和验证集中的诊断效能进行评价.结果从ADC-MRI图中共提取25508个影像组学特征,最终筛选出9个特征用于构建预测模型,这些特征在训练集和验证集中均与乳腺浸润性导管癌Luminal分型有显著相关性(P<0.01).该预测模型在训练集的曲线下面积(AUC)值为0.853[95%置信区间(CI)0.790~0.917],诊断敏感度为0.800,特异度为0.753,准确度为0.775,阳性预测值为0.641,阴性预测值为0.892;在验证集的AUC值为0.764(95%CI 0.664~0.864),诊断敏感度为0.800,特异度为0.712,准确度为0.738,阳性预测值为0.533,阴性预测值为0.897.结论基于ADC-MRI的影像组学预测模型可作为一种预测乳腺浸润性导管癌Luminal分型的有效方法.Objective To explore the value of apparent diffusion coefficient(ADC)-MRI based on radiomics in predicting the Luminal and non-Luminal breast invasive ductal carcinoma.Methods Pathologically confirmed breast invasive ductal carcinoma,241 patients were admitted.The clinical data and MRI images of those patients were retrospectively collected.According to the immunohistochemistry and fluorescence in situ hybridization,all patients were divided into four molecular subtypes:Luminal A,Luminal B,human epidermal growth factor receptor-2(HER-2)overexpression and basal-like.Luminal A and B were Luminal types,HER-2 overexpression and basal-like were non-Luminal types.All cases were randomly divided into training set(138 patients)and validation set(103 patients).Three-dimensional region of interest(3D-ROI)of lesions on the ADC-MRI map were sketched and radiomics features were extracted by computer automatically.regression was used to reduce dimensions,select features and build predictive model.The validation set was used to verify the prediction model.The diagnostic effectiveness of the prediction model in the training set and the validation set was evaluated using the receiver operating characteristic(ROC)curve.Results 25508 radiomics features were extracted from ADC-MRI map and 9 features were finally selected to build prediction models.The features were significantly correlated with Luminal types of breast invasive ductal carcinoma both in the training set and the validation set(P<0.01).The area under curve(AUC)of the predictive model in the training set was 0.853[95%confidence interval(CI)0.790-0.917],the diagnostic sensitivity was 0.800,the specificity was 0.753,the accuracy was 0.775,the positive predictive value was 0.641,and the negative predictive value was 0.892.The AUC in the validation set was 0.764(95%CI 0.664-0.864),the diagnostic sensitivity was 0.800,the specificity was 0.712,the accuracy was 0.738,the positive predictive value was 0.533 and the negative predictive value was 0.897.Conclusion The radiomics pr

关 键 词：乳腺浸润性导管癌 磁共振成像 影像组学 表观扩散系数 

分 类 号：R737.9[医药卫生—肿瘤] R.445.2[医药卫生—临床医学]