Wang Ailing, He Muzhen, Zhang Chengxiu, Zheng Yunyan, Song Yang, Wang Chenglong, Ma Mingping, Yang Guang
Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China.
Department of Radiology, Shengli Clinical Medical College of Fujian Medical University, Fujian Provincial Hospital, Fuzhou, China.
Front Oncol. 2025 Jun 27;15:1595157. doi: 10.3389/fonc.2025.1595157. eCollection 2025.
To explore the value of intravoxel incoherent motion (IVIM)-based habitat imaging in predicting immunohistochemistry in patients with breast cancer.
299 patients with suspected breast cancer were randomly assigned to a training set of 210 individuals and a test set of 89 individuals. A series of models was constructed for human epidermal growth factor receptor 2 (HER2)/Ki-67/hormone receptors (HR)/lymph node metastasis (LNM) prediction, including the whole-tumor model, habitat model, conventional MRI features (CF) model and hybrid model (incorporating habitats features and CF). The performance of various models was evaluated with the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). P (two-tailed) < 0.05 was considered statistically significant.
On the test cohort, for HER2/HR/LNM, the habitats model achieved the highest AUC values of 0.692/0.651/0.722, higher than those of the whole-tumor model (AUC = 0.591/0.599/0.609) and the CF model (AUC = 0.598/0.603/0.608). For Ki-67, the CF model achieved a highest AUC of 0.746. The hybrid model achieved AUC values of 0.706/0.762/0.668/0.728 for HER2/Ki67/HR/LNM. DeLong test showed a significant difference between habitats model and the whole-tumor model for LNM ( = 0.006).
While habitat features can provide richer biological information, the models combining habitats and CF obtained more accurate results than other models, making them promising candidates for clinical application in breast cancer diagnosis.
探讨基于体素内不相干运动（IVIM）的栖息地成像在预测乳腺癌患者免疫组化结果中的价值。
将299例疑似乳腺癌患者随机分为210例的训练集和89例的测试集。构建了一系列用于预测人表皮生长因子受体2（HER2）/ Ki-67/激素受体（HR）/淋巴结转移（LNM）的模型，包括全肿瘤模型、栖息地模型、传统MRI特征（CF）模型和混合模型（结合栖息地特征和CF）。采用受试者操作特征曲线（AUC）下面积和决策曲线分析（DCA）评估各模型的性能。P（双侧）<0.05被认为具有统计学意义。
在测试队列中，对于HER2/HR/LNM，栖息地模型的AUC值最高，分别为0.692/0.651/0.722，高于全肿瘤模型（AUC = 0.591/0.599/0.609）和CF模型（AUC = 0.598/0.603/0.608）。对于Ki-67，CF模型的最高AUC为0.746。混合模型对于HER2/Ki67/HR/LNM的AUC值分别为0.706/0.762/0.668/0.728。DeLong检验显示栖息地模型和全肿瘤模型在LNM方面存在显著差异（ = 0.006）。
虽然栖息地特征可以提供更丰富的生物学信息，但结合栖息地和CF的模型比其他模型获得了更准确的结果，使其有望用于乳腺癌诊断的临床应用。
