基于乳腺癌分子亚型的术前乳腺MRI特征研究  

Preoperative Breast MRI Features Based on Molecular Subtypes of Breast Cancer

作　　者：林霞 吴佳星 官桂林 左怀全 蒋富杰 LIN Xia;WU Jia-xing;GUAN Gui-lin;ZUO Huai-quan;JIANG Fu-jie(Department of Breast Surgery,the Affiliated Hospital of Southwest Medical University,Luzhou 646000,Sichuan Province,China;Department of Breast and thyroid Surgery,the First People's Hospital of Yibin,Yibin 641100,Sichuan Province,China;Chongqing Cancer research Institute,Yuzhong 410036,Chongqing,China)

机构地区：[1]西南医科大学附属医院乳腺外科,四川泸州646000 [2]宜宾市第一人民医院乳腺甲状腺外科,四川宜宾641100 [3]重庆市肿瘤研究所,重庆渝中410036

出　　处：《中国CT和MRI杂志》2025年第2期101-103,共3页Chinese Journal of CT and MRI

基　　金：重庆市自然科学基金面上项目(CSTB2022NSCQ-MSX1158)。

摘　　要：目的本研究旨在开发和验证基于动态对比增强磁共振成像(DCE-MRI)的放射组学特征,在浸润性乳腺癌患者术前区分管腔和非管腔分子亚型。方法将135例浸润性乳腺癌患者按7:3的比例分为训练组(n=95)和测试组(n=40),其中腔型(n=)78例,非腔型(n=57)。根据人口学和MRI影像特征构建临床危险因素。通过从DCE-MRI第二期图像中提取放射组学特征来构建放射组学特征,并计算放射组学得分(rad-core)。最后,从校正、区分和临床实用性方面对预测性能进行了评估。结果多因素Logistic回归分析显示,临床危险因素不是浸润性乳腺癌患者管腔和非管腔分子亚型的独立预测因素。同时,放射组学特征在训练集(AUC,0.86;95%CI,0.78~0.93)和测试集(AUC,0.80;95%CI,0.65~0.95)中表现出良好的区分性。结论DCE-MRI放射组学特征是一种在浸润性乳腺癌患者术前和非侵袭性区分腔分子亚型和非腔分子亚型的有前景的工具。Objective The purpose of this study is to develop and verify the radiohistological characteristics of dynamic contrast enhanced magnetic resonance imaging(DCE-MRI)to differentiate the lumen and non lumen molecular subtypes in patients with invasive breast cancer before surgery.Methods 135 patients with invasive breast cancer were divided into a training group(n=95)and a test group(n=40)according to the ratio of 7:3,including 78 cases of cavity type(n=)and non cavity type(n=57).Construct clinical risk factors based on demographic and MRI imaging features.Construct radiomics features by extracting radiomics features from the second phase of DCE-MRI images,and calculate the radiomics score(rad core).Finally,the predictive performance was evaluated in terms of calibration,differentiation,and clinical practicality.Results Multivariate logistic regression analysis showed that clinical risk factors were not independent predictors of lumen and non lumen molecular subtypes in patients with invasive breast cancer.Meanwhile,radiomics features showed good discriminability in the training set(AUC,0.86;95%CI,0.78-0.93)and the test set(AUC,0.80;95%CI,0.65-0.95).Conclusion The DCE-MRI radiological features are a promising tool for preoperative and non-invasive differentiation of lumen and non lumen molecular subtypes in patients with invasive breast cancer.

关 键 词：乳腺癌 分子亚型 乳腺MRI 动态对比增强磁共振成像 

分 类 号：R737.9[医药卫生—肿瘤]