基于合成MRI定量图谱的全肿瘤直方图模型预测浸润性导管乳腺癌腋窝淋巴结状态
本研究纳入2022年1月至10月212例浸润性导管癌患者,行乳腺MRI包括合成MRI检查,从T1-Gd等定量图谱及标准序列提取全肿瘤直方图特征。基于T1-Gd定量图谱特征(能量、最小值、方差)联合临床特征(年龄、多灶性)的模型在预测N0与N+组间腋窝淋巴结转移中AUC达0.879,合成T1-Gd模型单独预测AUC为0.823;合成T2_entropy与PD-Gd_energy区分N1与N2-3组AUC为0.722。结论认为合成MRI定量参数的全肿瘤直方图特征可作为术前无创预测腋窝淋巴结转移的补充方法。
要点
- T1-Gd定量图谱联合年龄、多灶性的模型预测N0与N+组腋窝淋巴结转移AUC为0.879。
- 合成T1-Gd定量图谱模型单独预测N0与N+组腋窝淋巴结转移AUC为0.823。
- 合成T2_entropy与PD-Gd_energy区分N1组与N2-3组AUC为0.722。
核心结论
- 合成MRI定量参数的全肿瘤直方图特征可作为术前无创预测腋窝淋巴结转移的补充方法。
研究设计与样本
- 研究类型为回顾性分析,纳入2022年1月至2022年10月共212例浸润性导管癌患者。
- 患者行乳腺MRI检查包括合成MRI,获得标准T2加权图像、DCE-MRI及合成MRI定量图谱。
- 从定量图谱、标准T2加权图像及DCE-MRI提取全肿瘤13个特征。
- 采用Student t检验、Mann-Whitney U检验、逻辑回归及ROC曲线进行统计分析。
- 同时评估基于常规2D感兴趣区获得的合成MRI定量参数均值。
诊断标准与关键数值
- 基于T1-Gd定量图谱特征(能量、最小值、方差)联合临床特征(年龄、多灶性)的联合模型预测N0组(无腋窝淋巴结转移)与N+组(转移腋窝淋巴结≥1枚)的AUC为0.879。
- 在单个定量图谱及标准序列衍生模型中,合成T1-Gd模型预测N0组与N+组的AUC为0.823。
- 合成T2_entropy与PD-Gd_energy区分N1组(转移腋窝淋巴结≥1枚且≤3枚)与N2-3组(转移腋窝淋巴结>3枚)的AUC为0.722。
临床应用与阅片要点
- 合成MRI定量图谱的全肿瘤直方图特征可用于术前无创评估腋窝淋巴结转移状态。
- T1-Gd定量图谱的能量、最小值、方差联合年龄、多灶性对N0与N+的鉴别诊断效能最高。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
PURPOSE: To investigate the potential of using histogram analysis of synthetic MRI (SyMRI) images before and after contrast enhancement to predict axillary lymph node (ALN) status in patients with invasive ductal carcinoma (IDC).
METHODS: From January 2022 to October 2022, a total of 212 patients with IDC underwent breast MRI examination including SyMRI. Standard T2 weight images, DCE-MRI and quantitative maps of SyMRI were obtained. 13 features of the entire tumor were extracted from these quantitative maps, standard T2 weight images and DCE-MRI. Statistical analyses, including Student's t-test, Mann-Whiney U test, logistic regression, and receiver operating characteristic (ROC) curves, were used to evaluate the data. The mean values of SyMRI quantitative parameters derived from the conventional 2D region of interest (ROI) were also evaluated.
RESULTS: The combined model based on T1-Gd quantitative map (energy, minimum, and variance) and clinical features (age and multifocality) achieved the best diagnostic performance in the prediction of ALN between N0 (with non-metastatic ALN) and N+ group (metastatic ALN ≥ 1) with the AUC of 0.879. Among individual quantitative maps and standard sequence-derived models, the synthetic T1-Gd model showed the best performance for the prediction of ALN between N0 and N+ groups (AUC = 0.823). Synthetic T2_entropy and PD-Gd_energy were useful for distinguishing N1 group (metastatic ALN ≥ 1 and ≤ 3) from the N2-3 group (metastatic ALN > 3) with an AUC of 0.722.
CONCLUSIONS: Whole-tumor histogram features derived from quantitative parameters of SyMRI can serve as a complementary noninvasive method for preoperatively predicting ALN metastases.
引用
- Zeng F, Yang Z, Tang X, et al. Whole-tumor histogram models based on quantitative maps from synthetic MRI for predicting axillary lymph node status in invasive ductal breast cancer. Eur J Radiol. 2024. PMID: 38262156. doi:10.1016/j.ejrad.2024.111325
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。