RESUMO
Objective To observe the value of radiomics models and nomogram model based on two-dimensional ultrasound and automated breast volume scanning(ABVS)for predicting molecular types of breast cancer.Methods Data of 326 female patients of single breast cancer confirmed by pathology were analyzed retrospectively.The patients were randomly divided into training set(n=260)or validation set(n=66)at the ratio of 8∶2,and further divided into Luminal subgroup and non-Luminal subgroup.Radiomics features were extracted based on two-dimensional ultrasound of breast and ABVS imaging,then model2DUS,modelABVS and modelcombined radiomics were constructed,respectively.Univariate analysis and multivariate logistic regression analysis were used to screen independent factors for predicting molecular types of breast cancer,and nomogram model(modelnomogram)was constructed combined with independent factors and radiomics Radscores.The receiver operating characteristic(ROC)curve was used to evaluate the efficacy of each model for molecular type of breast cancer.Results The maximum diameter of tumor(OR=1.029)and the retraction phenomenon(OR=0.408)were both independent predictive factors for molecular type of breast cancer(both P＜0.05).The area under the curve(AUC)of model2DUS,modelABVS.modelcombined radiomics and modelnomogram for predicting molecular type of breast cancer in validation set was 0.67,0.75,0.84 and 0.83,respectively.No significant difference of AUC of modelcombined radiomics and modelnomogram was found(P＞0.05),which were both higher than AUC of model2DUs and modelABVS(all P＜0.05).Conclusion Combined radiomics model and nomogram model based on two-dimensional ultrasound and ABVS could effectively predict molecular type of breast cancer.