基于DBT瘤内、瘤周影像组学的列线图对乳腺BI-RADS 4类病变良恶性的鉴别诊断价值研究OA
A Study on the nomogram based on DBT intra-tumoral and peri-tumoral radiomics for benign and malignant differentiation of breast BI-RADS 4 lesions
目的:探讨基于乳腺断层摄影(Digital breast tomosynthesis,DBT)瘤内、瘤周影像组学特征及临床影像学特征构建的列线图模型能否鉴别乳腺BI-RADS 4 类病变的良恶性.方法:回顾性分析 2021 年8 月—2024 年12 月南京医科大学第二附属医院190 例乳腺BI-RADS 4 类病变患者的DBT检查资料.按 7:3 比例随机分为训练集(n=133)和验证集(n=57).使用 3D-Slicer勾画瘤内感兴趣区(ROI),并向外周扩展 2mm获取瘤周ROI.使用Python提取影像组学特征,使用t检验、最小绝对收缩和选择法筛选最优特征,构建瘤内、瘤内+瘤周影像组学模型,计算影像组学评分(RadScore).使用Logistic回归筛选独立预测因子建立临床模型,结合RadScore构建列线图模型.使用受试者工作特征(ROC)曲线的曲线下面积(AUC)、校准曲线和决策曲线评估性能.结果:瘤内、瘤内+瘤周模型在训练集的AUC分别为0.808、0.856,在验证集的AUC分别为 0.702、0.830.Logistic回归显示肿块形态不规则、伴有肿大淋巴结和年龄是乳腺病变良恶性的独立预测因子(P<0.05).临床模型在训练集和验证集的AUC分别为 0.674、0.684.列线图模型在训练集和验证集的AUC分别为 0.900、0.872,性能优于单一模型(P<0.05).结论:基于DBT瘤内、瘤周影像组学特征及临床影像学特征的列线图模型能够较准确预测乳腺BI-RADS 4 类病灶的良恶性.
Objective:To investigate whether the nomogram model constructed based on intratumoral and peritumoral ra-diomic features of digital breast tomosynthesis(DBT)combined with clinical imaging features can distinguish benign from malig-nant breast BI-RADS category 4 lesions.Methods:A retrospective analysis was conducted on the DBT images of 190 female patients who underwent DBT at the Second Affiliated Hospital of Nanjing Medical University from August 2021 to December 2024.All patients were randomly assigned to a training group of 133 cases and a validation group of 57 cases in a 7:3 ratio.Delineate the intratumoral region of interest(ROI)using 3D-Slicer,and expand outward by 2 mm to obtain the peritumoral ROI.Radiomics features were extracted and analyzed using Python(Version 3.5).The t-test and least absolute shrinkage and selection operator(LASSO)method were used to screen the optimal features,and the intra-tumoral and peri-tumoral radiomics models were constructed.Then the radiomics signature's score(RadScore)was calculated.Logistic regression analysis was used to screen the independent risk factors and after that,a clinical model was constructed.The nomogram was constructed using RadScore and the independent risk factors.The differential diagnostic performance of the each model was evaluated using the area under the ROC curve(AUC),calibration and decision curves.Results:The AUC of intra-tumoral and intra-tumoral+peri-tumoral model was 0.808,0.856 in the training set and 0.702,0.830 in the validation set.Logistic regression analysis showed that irregular tumor morphology,enlarged lymph nodes and age were independent risk factors for benign and malignant differ-entiation of BI-RADS 4 lesions.The AUC values for the clinical model in the training and validation sets were 0.674 and 0.684,respectively.The AUC values of the nomogram model in the training set and validation set were 0.900 and 0.872,re-spectively,and its performance was superior to that of the single model(P<0.05).Conclusion:The nomogram model based on DBT intra-tumoral and peri-tumoral radiomics and clinical imaging features shows good performance in predicting the benign and malignant lesions of BI-RADS 4.
雍千叶;李海歌;马玉萍;王梅;郭浩东
南京医科大学第二附属医院医学影像科,江苏 南京 210011南京医科大学第二附属医院医学影像科,江苏 南京 210011南京医科大学第二附属医院医学影像科,江苏 南京 210011南京医科大学第二附属医院医学影像科,江苏 南京 210011南京医科大学第二附属医院医学影像科,江苏 南京 210011
医药卫生
乳腺肿瘤放射摄影术
Breast NeoplasmsRadiography
《中国临床医学影像杂志》 2026 (3)
173-177,5
南京医科大学科技发展基金项目(NMUB20230032).
评论