基于对比增强能谱X线摄影影像组学模型预测三阴性乳腺癌的价值

Study of the value of contrast-enhanced spectral mammography radiomics model for the prediction of triple-negative breast cancer

目的:探究基于乳腺对比增强能谱X线摄影(CESM)重建图像的影像组学模型预测三阴性乳腺癌的价值。方法:回顾性分析388例(训练集206例,测试集182例)乳腺癌患者的CESM图像,分别在头尾位、内外斜位重建图中勾画ROI并提取影像组学特征。在训练集中运用单变量分析及最小绝对值收敛和选择算子(LASSO)筛选特征并建立逻辑回归模型,运用ROC曲线与决策曲线对模型进行评价。结果:经筛选后模型包含12个特征。在测试集中模型的AUC为0.86[95%置信区间(95%CI)(0.80,0.92)],准确率、特异度及敏感度分别为0.77、0.77、0.76。在决策曲线分析中,诊断阈值<0.68时,模型具有较好的净收益。结论:基于CESM重建图像的影像组学模型对预测三阴性乳腺癌具有一定价值。

Objective:To explore the value of radiomics model based on contrast-enhanced spectral mammography(CESM)in the prediction of triple-negative breast cancer.Methods:CESM images of 388 patients with breast cancer(206 cases in training set and 182 cases in testing set)were retrospectively analyzed.ROIs were segmented on reconstructed imaging on cranial caudal(CC)and mediolateral oblique(MLO)and then radiomics features were extracted.Radiomics features were selected and univariate analysis,least absolute shrinkage and selection operator(LASSO)were used in the training set.The model was evaluated by ROC curve and decision curve.Results:A total of 12 radiomics features were included in radiomics model.AUC,accuracy,specificity and sensitivity of model in the testing set were 0.86[95%CI(0.80,0.92)],0.77,0.77 and 0.76,respectively.In decision curve analysis,if the threshold was less than 0.68,the model required good net benefit.Conclusion:CESM based radiomics model is valuable for the prediction of triple-negative breast cancer.