MRI影像组学3D及2D特征模型预测Luminal型乳腺癌Ki-67表达的价值  

The value of 3D and 2D radiomics features models of MRI in predicting Ki-67 expression in Luminal breast cancer

作　　者：尹阳[1] 李雯璐 郭济韬 张健 李娜[1] 赵艳[1] 杨志远[1] YIN Yang;LI Wenlu;GUO Jitao;ZHANG Jian;LI Na;ZHAO Yan;YANG Zhiyuan(Department of Radiology,the People’s Hospital of Leshan,Leshan,Sichuan Province 614000,China)

机构地区：[1]乐山市人民医院放射科,四川乐山614000

出　　处：《实用放射学杂志》2025年第1期52-57,共6页Journal of Practical Radiology

摘　　要：目的探讨基于多参数MRI影像组学3D及2D特征模型预测乳腺癌Ki-67表达状态(分别以14%和20%为临界值)的价值。方法回顾性分析经病理证实为Luminal型乳腺癌的147例患者的MRI图像,将患者按7︰3的比例随机分为训练集和测试集,从扩散加权成像(DWI)、动态对比增强(DCE)前蒙片(简称S0)、DCE第3期(简称S3)图像中提取瘤内联合瘤周的3D和2D影像组学特征,通过3种特征归一化方法、2种特征降维方法、4种特征选择方法、10种分类器进行多种流水线组合构建模型,通过受试者工作特征(ROC)曲线及曲线下面积(AUC)评估模型的预测效能,选出3D和2D单参数(DWI、S0、S3)及多参数组合(S0+S3、S0+DWI、S3+DWI、S0+S3+DWI)的最佳模型,通过DeLong检验比较不同模型间的差异。结果以14%为临界值,3D和2D各模型在训练集AUC为0.726~0.824、0.707~0.835,测试集AUC为0.724~0.82、0.701~0.805;以20%为临界值,3D和2D各模型在训练集AUC为0.743~0.868、0.793~0.881,测试集AUC为0.738~0.853、0.743~0.814;2个临界值标准中3D与2D相同参数模型间比较均无明显统计学差异。结论基于多参数MRI影像组学模型能够较好地预测乳腺癌Ki-67表达状态,基于瘤内联合瘤周的3D和2D特征模型具有同等预测效能。Objective To explore the value of 3D and 2D radiomics features models based on multiparameter MRI in predicting Ki-67 expression(with 14%and 20%as the critical values,respectively)in breast cancer.Methods The MRI images of 147 patients with pathologically confirmed Luminal breast cancer were analyzed retrospectively.The patients were randomly divided into training set and test set according to the ratio of 7︰3.The 3D and 2D radiomics features of intratumor and peritumor were extracted from diffusion weighted imaging(DWI),dynamic contrast enhancement(DCE)mask(S0)and DCE phase 3(S3)images.Then the models were constructed by multiple pipeline combinations of three feature normalization methods,two feature dimensionality reduction methods,four feature selection methods,and ten classifiers.The receiver operating characteristic(ROC)curve and the area under the curve(AUC)were used to evaluate the prediction performance of the models in order to select the best 3D and 2D single parameter(DWI,S0,S3)and multiparameter combination(S0+S3,S0+DWI,S3+DWI,S0+S3+DWI)models.Finally,the differences between the models were compared by DeLong test.Results With 14%as the critical value,the AUC of 3D and 2D models in the training set were 0.726-0.824 and 0.707-0.835,respectively,and those in the test set were 0.724-0.82 and 0.701-0.805.With 20%as the critical value,the AUC of 3D and 2D models in the training set were 0.743-0.868 and 0.793-0.881,respectively,and those in the test set were 0.738-0.853 and 0.743-0.814.There was no significant statistical difference between 3D and 2D models with the same parameter in the two critical values standards.Conclusion The multiparameter MRI-based radiomics models can better predict the expression of Ki-67 in breast cancer,and the models based on intratumor and peritumor 3D and 2D features have the same prediction efficiency.

关 键 词：影像组学 乳腺癌 KI-67 磁共振成像 

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