DCE-MRI Based Radiomics Signature For Preoperatively Predicting Histological Grade Of Breast Invasive Ductal Carcinoma

Objective: To explore the relationship between radiomics features based on Dynamic Contrast Enhanced-Magnetic resonance imaging(DCE-MRI)and histological grade of breast invasive ductal carcinoma(IDC).To develop and validate radiomics signature based on DCE-MRI to preoperatively evaluate the histological grade(grade I,grade II,grade III)of breast IDC,and to evaluate the diagnostic performance of radiomics signature.Methods:This retrospective study included 201 patients of IDC in our hospital from April 2013 to December 2019,all patients received preoperative breast MRI and confirmed by pathological diagnosis,further were divided into training group 121cases(31 cases of grade I,46 cases of grade II,44 cases of grade III)and validation group80 cases(21 cases of grade I,30 cases of grade II,29 cases of grade III).After the ROIs were manually delineated around the tumor profile layer by layer,radiomics features were extracted by pyradiomics,valuable features were selected by statistical methods and the least absolute shrinkage and selection operator(LASSO)-logistic regression model,radiomics signature was build by regression coefficient of linear combination screening feature.Kruskal-Wallis Test was used to explore the difference of different histological grade of breast IDC groups with radiomics features values and radiomics signature values.Receiver operating characteristics(ROC)curve was applied to evaluate the diagnostic performance of the radiomics signature in training group with area under curve(AUC),sensitivity,specificity,and accuracy,the validation group dataset was used to verify the predition performancer of radiomics signature to preoperatively evaluate the histological grade of breast IDC.Results:Finally,nine radiomics features were selected,the differences of the nine radiomics features values in the different histological grade groups of breast IDC were statistically significant(P<0.05).Nine radiomics features were selected to build a radiomics signature,which showed good performance for preoperatively evaluating histological grade of breast IDC,with AUC of 0.776,sensitivity of 63.6%,specificity of 80.3%and accuracy of 69.3%in training group dataset and AUC of 0.728,sensitivity of 68.3%,specificity of 67.4%and accuracy of 68.0%in the validation group dataset.Conclusion: radiomics features based on DCE-MRI were helpful to distinguish the histological grade of breast IDC;the radiomics signature based on DCE-MRI has certain value to preoperatively evaluate the histological grade of breast IDC,which may provide valuable clinical information.