机构地区:[1]新疆医科大学第三临床医学院,新疆乌鲁木齐830011 [2]新疆医科大学附属肿瘤医院介入诊疗科,新疆乌鲁木齐830011 [3]新疆医科大学附属肿瘤医院超声科,新疆乌鲁木齐830011 [4]南方医科大学第十附属医院超声科,广东东莞523000
出 处:《分子影像学杂志》2025年第1期1-9,共9页Journal of Molecular Imaging
基 金:国家自然科学基金(82360362)。
摘 要:目的基于超声瘤内、瘤周影像组学与临床超声特征联合构建的列线图模型在三阴性乳腺癌中的预测价值。方法回顾性收集2018年1月~2023年12月新疆医科大学附属肿瘤医院540例经病理证实为浸润性乳腺癌的患者的相关资料,根据免疫组织化学检查结果分为TNBC组(n=263)和非TNBC组(n=277),按7∶3随机分为训练集(n=378)和验证集(n=162)。采用Pyradiomics在原始图像及衍生图像的瘤内、瘤周ROI提取影像组学特征;通过Pearson相关系数、最大相关最小冗余(mRMR)和最小绝对收缩和选择算子(LASSO)筛选特征;分别建立瘤内、瘤周、瘤内+瘤周模型并获得影像组学标签;经单因素、多因素Logistic回归分析从临床超声特征中筛选TNBC的独立预测因子并构建临床模型;选择效能最佳的影像组学标签与临床独立预测因子构建列线图模型。绘制ROC曲线、校准曲线及临床决策曲线评价模型效能,通过Delong检验比较模型曲线下面积(AUC)的差异。结果最终通过瘤内(n=14)、瘤周(n=15)、瘤内+瘤周(n=19)影像组学特征构建模型,以瘤内+瘤周影像组学模型预测效能最佳,训练集和验证集AUC分别为0.941、0.934。将肿块大小、乳腺超声背景、微钙化、后方回声作为独立预测因子构建临床模型,训练集和验证集AUC为0.906、0.911。瘤内+瘤周影像组学标签联合肿块大小、乳腺超声背景、微钙化、后方回声构建列线图模型,训练集和验证集AUC分别提升至0.974、0.965。结论基于超声瘤内、瘤周影像组学与临床超声特征联合构建的列线图模型在预测三阴性乳腺癌中具有较高价值,可以为临床决策提供借鉴和参考。Objective To evaluate the predictive value of the nomogram model constructed based on the combination of ultrasound intratumor and peritumor radiomics and clinical ultrasound features in triple-negative breast cancer(TNBC).Methods The relevant data of 540 patients with pathologically confirmed invasive breast cancer in the Affiliated Cancer Hospital of Xinjiang Medical University were retrospectively collected from January 2018 to December 2023,and divided into the TNBC group(n=263)and the non-TNBC group(n=277)according to the results of immunohistochemistry,and randomly divided into the training set(n=378)and the validation set(n=162)according to the 7:3 randomization.The radiomics features were extracted using Pyradiomics in the intratumoral and peritumoral ROIs of the original and derived images;the features were screened by Pearson's correlation coefficient,maximal correlation minimal redundancy and LASSO.The intratumoral,peritumoral,and intratumoral+peritumoral models were established and radiomics signature were obtained,respectively,and after univariate and multivariate Logistic regression analysis was used to screen the independent predictors of TNBC from clinical ultrasound features and construct the clinical model.The radiomics signature with the best efficacy and the clinical independent predictors were selected to construct the nomogram model.ROC curves,calibration curves and clinical decision curves were plotted to evaluate the model efficacy,and the differences in the area under the curve(AUC)of the models were compared by the Delong test.Results The models were finally constructed by intratumoral(n=14),peritumoral(n=15),and intratumoral+peritumoral(n=19)radiomics features,and the intratumoral+peritumoral radiomics model had the best predictive efficacy,with AUCs of 0.941 and 0.934 for the training and validation sets,respectively.The clinical models were constructed by using the tumor size,background echotexture,microcalcifications,and posterior echoes as independent predictors,and the AUCs of trai
分 类 号:R445.1[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...