Cao Mingtai, Liu Xinyi, Yang Airu, Xu Yuan, Zhang Qian, Cao Yuntai
Department of Radiology, Affiliated Hospital of Qinghai University, Xining, China.
Department of Radiology, The Second Hospital & Clinical Medical School, Lanzhou University, Lanzhou, China.
Magn Reson Imaging. 2025 Oct;122:110434. doi: 10.1016/j.mri.2025.110434. Epub 2025 Jun 1.
This study aims to explore the value of multiparametric magnetic resonance imaging (MRI) techniques-dynamic contrast-enhanced MRI (DCE-MRI), diffusion-weighted imaging (DWI), and T2-weighted fat-suppressed imaging (T2WI)-in predicting human epidermal growth factor receptor 2 (HER-2) status in breast cancer by integrating intratumoral and peritumoral radiomics features to establish a multiparametric MRI intratumoral and peritumoral radiomics model.
A retrospective cohort of 266 female breast cancer patients was analyzed. Patients from Center 1 (n = 199) were divided into a training set (n = 140) and internal validation set (n = 59; 7:3 ratio), while Center 2 (n = 67) provided the external test set. Using 3D Slicer, tumor boundaries were manually segmented on T2WI, DWI, and DCE-MRI to define intratumoral volumes of interest (VOIs). These VOIs were expanded by 3 mm to generate peritumoral regions (VOI_Peri3mm). Radiomics features were extracted from both regions, optimized via feature selection, and used to train eight random forest (RF) models. Performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA).
The multiparametric MRI intratumoral and peritumoral radiomics model (DWI_Peri3 + T2WI_Peri3 + DCE_Peri3_RF) demonstrated optimal HER-2 prediction, achieving area under the curve (AUC) values of 0.822 (95 % CI:0.755-0.889), 0.823 (0.714-0.932), and 0.813 (0.712-0.914) in the training, internal validation, and external test sets, respectively. It significantly outperformed single-parameter or single-region models and maintained cross-cohort consistency.
The intratumoral-peritumoral radiomics fusion model integrating DWI, T2WI, and DCE-MRI provides high diagnostic accuracy for HER-2 assessment, offering non-invasive biomarkers and enhancing precision in breast cancer management.
本研究旨在通过整合肿瘤内和肿瘤周围的影像组学特征，探索多参数磁共振成像（MRI）技术——动态对比增强MRI（DCE-MRI）、扩散加权成像（DWI）和T2加权脂肪抑制成像（T2WI）在预测乳腺癌人表皮生长因子受体2（HER-2）状态中的价值，以建立多参数MRI肿瘤内和肿瘤周围影像组学模型。
分析了266例女性乳腺癌患者的回顾性队列。来自中心1的患者（n = 199）被分为训练集（n = 140）和内部验证集（n = 59；7:3比例），而中心2的患者（n = 67）提供外部测试集。使用3D Slicer在T2WI、DWI和DCE-MRI上手动分割肿瘤边界，以定义肿瘤内感兴趣体积（VOI）。这些VOI扩大3mm以生成肿瘤周围区域（VOI_Peri3mm）。从两个区域提取影像组学特征，通过特征选择进行优化，并用于训练八个随机森林（RF）模型。使用受试者操作特征（ROC）曲线、校准曲线和决策曲线分析（DCA）评估性能。
多参数MRI肿瘤内和肿瘤周围影像组学模型（DWI_Peri3 + T2WI_Peri3 + DCE_Peri3_RF）显示出最佳的HER-2预测能力，在训练集、内部验证集和外部测试集中的曲线下面积（AUC）值分别为0.822（95%CI：0.755-0.889）、0.823（0.714-0.932）和0.813（0.712-0.914）。它显著优于单参数或单区域模型，并保持了跨队列的一致性。
整合DWI、T2WI和DCE-MRI的肿瘤内-肿瘤周围影像组学融合模型为HER-2评估提供了高诊断准确性，提供了非侵入性生物标志物，并提高了乳腺癌管理的精准性。
