基于DCE-MRI影像组学的三维特征术前预测乳腺癌Ki-67的表达水平  

Preoperative prediction of Ki-67 expression level in breast cancer based on three-dimensional features of DCE-MRI imaging

作　　者：林进丽 邓顺坚 方立广 张慈慈 周全 LIN Jinli;DENG Shunjian;FANG Liguang;ZHANG Cici;ZHOU Quan(Department of Medical Imaging,The Third Affiliated Hospital of Southern Medical University,Guangzhou 510630,Guangdong,China;Department of Medical Imaging,The Fifth Affiliated Hospital of Guangzhou Medical University,Guangzhou 510799,Guangdong,China;Department of Medical Imaging,Guangzhou Red Cross Hospital,Guangzhou 510240,Guangdong,China)

机构地区：[1]南方医科大学附属第三医院医学影像科,广东广州510630 [2]广州医科大学附属第五医院医学影像科,广东广州510799 [3]广州市红十字会医院医学影像科,广东广州510240

出　　处：《暨南大学学报（自然科学与医学版）》2024年第2期150-159,共10页Journal of Jinan University(Natural Science & Medicine Edition)

基　　金：国家自然科学基金重点项目(81471659)。

摘　　要：目的:构建基于动态增强磁共振成像(DCE-MRI)的影像组学三维特征模型,旨在术前预测乳腺癌Ki-67表达水平。方法:回顾性分析110例浸润性乳腺癌女性患者的临床、病理及DCE-MRI资料。根据细胞核增殖指数(Ki-67)将患者分为两组,Ki-67≥30%者为高表达组,Ki-67<30%者为低表达组。提取DCE-MRI第4期肿瘤区域三维影像组学特征;筛选出最优特征并以此构建预测乳腺癌Ki-67的表达状态模型,绘制诺莫图对模型可视化,并对模型效能进行评估。使用自助重抽样法(Bootstrap)重复抽样1000次,重新生成训练样本并重构模型进行内部验证。结果:成功构建模型的诺莫图,训练组的受试者工作特征(ROC)显示曲线下的面积(AUC)为0.876(95%CI为0.803~0.949)、最佳截断值为0.513、敏感度为80.6%、特异度为86.9%。Hosmer-Lemeshow拟合优度检验为0.735。决策曲线(DCA)阈值为17%~100%。内部验证组中AUC为0.854(95%CI为0.851~0.878)、敏感度为84.0%、特异度为72.9%。结论:基于DCE-MRI影像组学三维特征可在术前对乳腺癌Ki-67的表达状态进行预测,提供了一种能够在术前进行有效、无创性评估乳腺癌肿瘤细胞增殖状况的影像组学模型,为乳腺癌的治疗决策选择提供新的参考依据。Objective:To construct a 3D imaging model based on dynamic enhanced magnetic resonance imaging(DCE-MRI)to predict the expression level of Ki-67 in breast cancer.Methods:The clinical,pathological and MR Dynamic enhanced imaging data of 110 female patients with invasive breast cancer were analyzed retrospectively.Patients were divided into two groups according to Ki-67 value,those with Ki-67 value≥30%was defined as high expression group and<30%was defined as low expression group.The 3D imaging features of the tumor area in the fourth stage of DCE-MRI were extracted.The optimal features were selected to construct the expression state model of Ki-67 prediction in breast cancer,and the Nomogram was drawn to visualize the model and evaluate the effectiveness of the model.Bootstrap was used to sample the training sample for 1000 times and reconstruct the model for internal verification.Results:The receiver operating characteristic(ROC)of the training group showed that the area under the curve(AUC)was 0.876(95%CI 0.803~0.949),the optimal cut-off value was 0.513,the sensitivity was 80.6%,and the specificity was 86.9%.The Hosmer-Lemeshow goodness of fit test was 0.735.The decision curve(DCA)threshold ranges from 17%to 100%.In the internal validation group,AUC was 0.854(95%CI 0.851~0.878),sensitivity was 84.0%,and specificity was 72.9%.Conclusion:The expression status of Ki-67 in breast cancer can be predicted before surgery based on the 3D features of DCE-MRI imaging,which provides an effective and non-invasive imaging model to evaluate the proliferation of breast cancer tumor cells before surgery,and provides a new reference for the treatment decision selection of breast cancer.

关 键 词：影像组学 乳腺癌 细胞核增殖指数(Ki-67) 诺莫图 

分 类 号：R4[医药卫生—临床医学]