磁共振成像2025，Vol.16Issue(10)：41-47,97,8.DOI:10.12015/issn.1674-8034.2025.10.007
基于多参数MRI影像组学联合MRI特征无创预测NME型乳腺癌HER-2过表达和低表达
Non-invasive prediction of HER-2 overexpression and low expression in NME-type breast cancer using multiparametric MRI radiomics combined with MRI features
摘要
Abstract
Objective:To explore the value of multiparametric MRI radiomics combined with MRI features in non-invasively predicting human epidermal growth factor receptor 2(HER-2)overexpression and low expression in non-mass enhancement(NME)-type breast cancer.Materials and Methods:A total of 156 breast cancer cases with NME on dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)and pathologically confirmed were collected from our hospital,and divided into the HER-2 overexpression group(66 cases)and the HER-2 low expression group(90 cases).They were randomly assigned to a training set(124 cases)and a test set(32 cases)at a ratio of 8∶2.Volumes of interest(VOIs)were segmented on the 2nd phase(DCE-2),8th phase(DCE-8)of DCE-MRI,and diffusion weighted imaging(DWI)sequences,and radiomic features were extracted.The Elastic Net(Enet)algorithm was used to construct models based on DCE-2,DCE-8,DWI,and their combination.Logistic regression analysis was performed to identify independent influencing factors for HER-2 expression.Finally,a fusion model was built by combining the rad-score of the combined model with independent influencing factors.Results:The areas under the curve(AUC)of the radiomic models based on DCE-2,DCE-8,DWI,and their combination in the training and test sets were 0.746 and 0.714,0.768 and 0.714,0.721 and 0.635,0.823 and 0.734,respectively.Logistic regression analysis showed that the maximum tumor diameter was an independent factor for distinguishing HER-2 expression(P<0.05).The fusion model achieved the best predictive performance,with AUCs of 0.844 and 0.808 in the training and test sets,respectively.DeLong's test indicated no significant difference between the combined model and the fusion model(P=0.316).Analysis of SHAP results showed that rad-score contributed the most to the fusion model.Conclusions:Multi-parametric MRI radiomics combined with MRI features can effectively predict HER-2 overexpression and low expression in NME-type breast cancer,and the combination with SHAP algorithm can further improve the interpretability of the model.关键词
非肿块强化型乳腺癌/对比增强磁共振成像/影像组学/磁共振成像/人表皮生长因子受体2/过表达和低表达Key words
non-mass enhancement breast cancer/dynamic contrast-enhanced magnetic resonance imaging/radiomics/magnetic resonance imaging/human epidermal growth factor receptor 2/overexpression and low expression分类
医药卫生引用本文复制引用
赵盈,蒋鑫垚,赵楠,许永生,雷军强..基于多参数MRI影像组学联合MRI特征无创预测NME型乳腺癌HER-2过表达和低表达[J].磁共振成像,2025,16(10):41-47,97,8.基金项目
The Provincial Key Talent Project of the Organization Department of the Gansu Provincial CPC Committee in 2023(No.2023RCXM06). 2023年甘肃省委组织部省级重点人才项目(编号:2023RCXM06) （No.2023RCXM06）
