基于生境成像技术的动态增强MRI影像组学特征预测乳腺癌新辅助治疗病理完全缓解的价值  

The value of dynamic enhanced MRI radiomics features based on habitat imaging technology for predicting pathological complete remission in neoadjuvant treatment of breast cancer

作　　者：宋德领 闻彩云[1] 邰云鹏[1] 刘瑾瑾[1] 王美豪[1] 曹国全[1] Song Deling;Wen Caiyun;Tai Yunpeng;Liu Jinjin;Wang Meihao;Cao Guoquan(Department of Radiology,the First Affiliated Hospital of Wenzhou Medical University,Wenzhou 325000,China)

机构地区：[1]温州医科大学附属第一医院放射科,温州325000

出　　处：《中华放射学杂志》2025年第4期401-408,共8页Chinese Journal of Radiology

基　　金：浙江省教育厅课题项目(Y202457145);浙江省医药卫生科研基金项目(2025KY1005)。

摘　　要：目的基于生境成像技术探讨动态对比增强(DCE)-MRI影像组学特征对乳腺癌新辅助治疗(NAT)病理完全缓解(pCR)的预测价值。方法该研究为横断面研究。回顾性收集2016年8月至2023年12月在温州医科大学附属第一医院经病理证实为原发浸润性乳腺癌的119例患者的临床病理和影像资料。患者均为女性,年龄25~67岁。采用随机分层抽样方法按7∶3的比例分为训练集(n=83)和验证集(n=36)。根据Miller-Payne分级系统将患者分为病理完全缓解(pCR)与非病理完全缓解(non-pCR)。所有患者在NAT前均接受DCE-MRI,采用ITK-Snap软件勾画兴趣区(ROI),提取并筛选整个肿瘤区域影像组学特征并构建预测NAT后pCR传统影像组学模型(ROI_(整体)模型);利用生境成像技术将肿瘤区域分为3个亚区,提取并筛选ROI亚区1、ROI_(亚区)2、ROI_(亚区)3内的影像组学特征,并构建预测NAT后pCR的生境成像模型(ROI亚区1模型、ROI_(亚区)2模型、ROI_(亚区)3模型)。采用单因素logistic回归法筛选独立预测NAT后pCR的临床因素并构建临床预测模型。纳入临床预测模型和生境模型建立联合预测模型。采用受试者操作特征曲线和曲线下面积(AUC)评价各模型预测乳腺癌NAT后pCR的效能,采用决策分析曲线评价模型的临床应用效能。结果119例患者中,pCR患者74例,其中训练集52例,验证集22例,non-pCR患者45例,其中训练集31例,验证集14例。单因素logistic回归分析显示人表皮生长因子受体2状态(OR=0.254,95%CI 0.093~0.697,P=0.008)是NAT后pCR的独立预测因素,以此构建临床预测模型。训练集和验证集中ROI亚区1模型、ROI亚区2模型预测NAT后pCR的效能高于传统影像组学模型(ROI整体模型),训练集中AUC分别为0.805、0.748和0.728;验证集中AUC分别为0.776、0.718和0.708。训练集和验证集中,联合预测模型预测乳腺癌NAT后pCR效能AUC分别为0.877和0.818。决策分析曲线显示联合预测模型的净�ObjectiveTo investigate the predictive value of radiomics features derived from dynamic contrast-enhanced MRI(DCE-MRI)based on habitat imaging technology for pathological complete response after neoadjuvant therapy(NAT)for breast cancer.MethodsAll patients were female,aged 25-67 years.Patients were stratified into training(n=83)and validation(n=36)sets via stratified random sampling(7∶3 ratio).Pathological complete remission(pCR)and non-pathological complete remission(non-pCR)were defined using the Miller-Payne grading system.All patients underwent DCE-MRI before NAT.ITK-Snap software was used to outline the region of interest(ROI),the imaging histological features of the entire tumor region were extracted and screened,a traditional imaging histological model for predicting post-NAT pCR(ROI_(overall)model)was constructed;the tumor region was divided into three subregions using habitat imaging technology,and the imaging histological features within ROI_(subregion 1),ROI_(subregion 2),and ROI_(subregion 3)were extracted and screened,and the habitat imaging model for predicting post-NAT pCR were constructed(ROI_(subregion 1)model,ROI_(subregion 2)model,ROI_(subregion 3)model).Univariate logistic regression identified clinical predictors of pCR for clinical model construction.Combined models integrating clinical predictors and habitat imaging features were established.The efficacy of each model in predicting pCR after NAT in breast cancer was evaluated using receiver operating characteristic curves and area under the curve(AUC),and the efficacy of clinical application of the models was evaluated using decision curve analysis(DCA).ResultsOf the 119 patients,74 were pCR patients,with 52 in the training set and 22 in the validation set,and 45 were non-pCR patients,with 31 in the training set and 14 in the validation set.Logistic regression analysis showed that human epidermal growth factor receptor 2 status(OR=0.254,95%CI 0.093-0.697,P=0.008)was an independent predictor of pCR after NAT,and this was used to construct

关 键 词：乳腺肿瘤 磁共振成像 影像组学 新辅助治疗 生境成像 

分 类 号：R737.9[医药卫生—肿瘤]