背景：有一半的乳腺癌表达人表皮生长因子受体2（HER2）的水平较低，并可以通过新的抗体药物结合物进行靶向治疗。HER2零（免疫组化学[IHC]得分为0）、HER2低（IHC得分为1+或2+，并在荧光原位杂交[ FISH ]中存在阴性结果）和HER2阳性（IHC得分为2+，并在FISH或IHC得分为3+时显示阳性结果）乳腺癌之间的成像差异尚不清楚。 目的：评估多参数动态增强磁共振成像（MRI）基于放射组学特征是否能够帮助区分乳腺癌中的HER2表达。 材料和方法：本研究纳入了2020年12月至2022年12月间，在两个不同的中心接受MRI检查的乳腺癌女性患者。在T2加权和动态增强T1加权图像上进行肿瘤分割和放射组学特征提取。然后使用无监督相关性分析和最小绝对收缩选择算法来选择特征以构建放射组学签名。采用受试者工作特征曲线下面积（AUC）评估放射组学签名的性能。使用多变量Logistic回归分析来确定区分HER2表达的独立预测因子，包括训练集和预先获取的外部数据集。 结果：训练集包括来自中心1的208名患者（平均年龄53岁±14 [标准差]），外部测试集包括来自中心2的131名患者（平均年龄54岁±13）。在外部测试数据集中，放射组学签名在区分HER2低和阳性肿瘤与HER2零肿瘤方面的AUC为0.80 （95% CI: 0.71, 0.89），并且是区分这两组的重要预测因子（OR = 7.6；95% CI: 2.9, 19.8；P < .001）。在HER2低或阳性乳腺癌中，组织学类型、相关的非肿块性增强以及MRI上的多个病灶在外部测试集中对预测HER2阳性与HER2低乳腺癌的AUC为0.77（95% CI: 0.68, 0.86）。 结论：来自多参数乳腺MRI的放射组学签名和肿瘤描述特征可能预测乳腺癌中不同的HER2表达，并具有治疗意义。© RSNA，2023 本文的补充资料可在此文章上找到。此外，请参阅本期社论By Kataoka and Honda。

Background Half of breast cancers exhibit low expression levels of human epidermal growth factor receptor 2 (HER2) and can be targeted by new antibody-drug conjugates. The imaging differences between HER2-zero (immunohistochemistry [IHC] score of 0), HER2-low (IHC score of 1+ or 2+ with negative findings at fluorescence in situ hybridization [FISH]), and HER2-positive (IHC score of 2+ with positive findings at FISH or IHC score of 3+) breast cancers were unknown. Purpose To assess whether multiparametric dynamic contrast-enhanced MRI-based radiomic features can help distinguish HER2 expressions in breast cancer. Materials and Methods This study included women with breast cancer who underwent MRI at two different centers between December 2020 and December 2022. Tumor segmentation and radiomic feature extraction were performed on T2-weighted and dynamic contrast-enhanced T1-weighted images. Unsupervised correlation analysis of reproducible features and least absolute shrinkage and selector operation were used for the selection of features to build a radiomics signature. The area under the receiver operating characteristic curve (AUC) was used to assess the performance of the radiomic signature. Multivariable logistic regression was used to identify independent predictors for distinguishing HER2 expressions in both the training and prospectively acquired external data set. Results The training set included 208 patients from center 1 (mean age, 53 years ± 14 [SD]), and the external test set included 131 patients from center 2 (mean age, 54 years ± 13). In the external test data set, the radiomic signature achieved an AUC of 0.80 (95% CI: 0.71, 0.89) for distinguishing HER2-low and -positive tumors versus HER2-zero tumors and was a significant predictive factor for distinguishing these two groups (odds ratio = 7.6; 95% CI: 2.9, 19.8; P < .001). Among HER2-low or -positive breast cancers, histology type, associated nonmass enhancement, and multiple lesions at MRI had an AUC of 0.77 (95% CI: 0.68, 0.86) in the external test set for the prediction of HER2-positive versus HER2-low cancers. Conclusion The radiomic signature and tumor descriptors from multiparametric breast MRI may predict distinct HER2 expressions of breast cancers with therapeutic implications. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Kataoka and Honda in this issue.