The Value Of Preoperative Prediction Of Histological Grade Of Invasive Ductal Carcinoma Of The Breast Based On ADC And DCE Maps And Semi-quantitative Parameters Of MRI To Construct A Radiomics Model

ObjectiveTo investigate the value of constructing imaging radiomics model based on ADC,DCE map and semi-quantitative parameters of MRI for preoperative prediction of radiomics grading of invasive ductal carcinoma of breast for clinical application.Methods(1)Retrospective analysis of 114 patients with IDC breast cancer who underwent MRI with complete data between January 2020 and April 2022;58 cases were classified as low grade(grade Ⅰ+Ⅱ)and 56 cases as high grade(grade Ⅲ)according to histological grading.This study uses bootstrap for internal validation,so there is no need to divide the sample size into a training set and a validation set.(2)The age and gender of the patients included in the study were collected,as well as visual radiological characteristics including:enhancement pattern of the mass,borders,and number of lymph node metastases;and semi-quantitative parameters of MRI including:largest diameter,ADC value,early intensive rate,Time to max,and Time Intensity Curve Manual were obtained by post-processing;and the histological grading of the patients’ pathology was collected.(3)Using 3D-Slicer 5.0.3 to outline the region of interest layer by layer on ADC and DCE images respectively,687 and 617 image radiomics features were extracted;then Intraclass Correlation Efficient,Spearman correlation analysis and LASSO were used to sequentially The descending and screening were performed.Univariate,multifactor analysis and logistic regression analysis were used for screening of visual radiological features and MRI semi-quantitative parameters.Logistic regression analysis was used to construct 3 models,for ADC radiomics model,DCE radiomics model and ADC+DCE radiomics model.The best imaging radiomics features were selected to combine with the visual radiological features and MRI semi-quantitative parameters screened out as predictors to construct a comprehensive model.(4)ROC,AUC,Delong test,Brier Score,calibration curve and HosmerLemeshow test were used for model performance evaluation.Results(1)The AUC values of the radiomics model predicting histological grading of IDC breast cancer were 0.668(95%CI:0.5676-0.768),0.695(95%CI:0.599-0.7908)and 0.723(95%CI:0.6307-0.8145)for the ADC,DCE and ADC+DCE groups,respectively;the BS scores were 0.227,0.217,and 0.209,respectively,all<0.25.(2)The EER value was screened as a predictor by single-and multi-factor analysis and logistic regression analysis of visual radiological characteristics and MRI semi-quantitative parameters,and the best model was selected from the above three imaging radiomics models and combined with the EER to construct a comprehensive model;the AUC value of this model for predicting the histological grade of IDC breast cancer was 0.742(95%CI:0.6532-0.8314)and the BS score was 0.203<0.25.The best prediction efficiency of the integrated imaging model was obtained,and the calibration curve showed good agreement between the predicted and actual probabilities.ConclusionThe integrated imaging radiomics model constructed based on the ADC,DCE maps and EER values of MRI and its Nomogram have clinical value in predicting the histological grading of breast cancer in IDC and are expected to help clinicians optimize preoperative decision making.