The Value Of Ultrasound-based Radiomics Nomogram In Predicting The Pathological Classification Of Breast Fibroepithelial Neoplasms

Objective:A nomogram model based on ultrasound radiomics and clinical independent predictors was developed and validated to predict the pathological classification of breast fibroepithelial neoplasms and improve the correct diagnosis rate of breast phyllodes tumors(PT).Methods:The retrospective study included 106 breast PT patients and 220 fibroadenoma(FA)patients,which were randomly divided into training set and validation set.Analyze the age and ultrasound images features of breast lesions in the PT group and FA group.849 ultrasound-based radiomics features were extracted from gray-scale ultrasound images.The training set was subjected to univariate analysis,Least absolute shrinkage and selection operator(LASSO)algorithm and 10-fold cross validation for feature selection,and 14 radiomics features were selected.Further the radiomics score of each lesion is calculated.The radiomics models based on radiomics scores,clinical models based on clinical independent predictors and nomogram models based on radiomics scores combined with clinical independent predictors are constructed respectively,and verify them in the validation set.ROC curve was used to compare the predictive power of three modles for classification of breast fibroepithelial neoplasms.Results:There were significant differences between the two groups in age and maximum diameter,internal echo,morphology,edge,rear echo enhancement,and blood flow abundance of the lesion(P<0.05).Age,the maximum diameter and internal echo of the lesion are clinical independent predictors.The area under the ROC curve AUC of the radiomics model,clinical model and nomogram model in the training set were 0.836(95%CI,0.782～0.890),0.835(95%CI,0.775～0.894),0.869(95%CI,0.820～0.918),and the AUC in the validation set were 0.807(95%CI,0.712～0.902),0.820(95%CI,0.730～0.909),0.856(95%CI,0.775～0.936),respectively.In the pairwise comparison of the three prediction models,the AUC of the nomogram model was higher than that of the clinical model,and the difference was statistically significant(P<0.05).Conclusions:Ultrasound-based radiomics model has a certain value in predicting the classification of breast fibroepithelial neoplasms.The ultrasound-based radiomics nomogram model combined with clinical parameters will help to further improve the prediction efficiency.Ultrasound-based radiomics may become a potential bioimaging marker for predicting the pathological classification of breast fibroepithelial neoplasms.