Prediction of Histological Grade in Invasive Ductal Carcinoma of the Breast Using Multi-parameter MRI Radiomic Features

LI Gang; LV Xiaohong; ZHENG Lanqing; ZHANG Fan; WANG Danni; LI Xiao; TONG Yuezhu; LIU Ning

doi:10.3969/j.issn.2096-305X.2026.01.015

Abstract：Objective To evaluate the predictive efficacy of multiparametric dynamic contrast-enhanced MRI(DCE-MRI)radiomics for histological grading in invasive ductal carcinoma(IDC)of the breast.Methods A total of 183 cases of IDC confirmed by pathology the First Affiliated Hospital of Jinzhou Medical University were included(139 cases of grade Ⅰ-Ⅱ and 44 cases of grade Ⅲ),and they were divided into training and test sets at a ratio of 7:3.Preprocessing,3D segmentation and feature extraction were conducted on T2WI,DWI,ADC and enhanced S1-S5 phases;features were selected through t-test,LASSO and variance inflation factor,and random forest(RF),support vector machine(SVM),logistic regression(LR),naive Bayes(NB)and multi-model fusion(RF+SVM+LR)were constructed.Receiver operating characteristic(ROC)curves were used to calculate the area under the curve(AUC),with DeLong's test comparing model performance.Decision curve analysis(DCA)and calibration curves assessed clinical applicability.Results The S3 phase(third post-contrast phase)demonstrated optimal performance among single-sequence models;in the testing set,RF,SVM,LR,and NB models trained on S3-phase features achieved AUCs of approximately 0.81,0.84,0.85,and 0.75,respectively;the multi-model fusion(S3 phase)achieved an AUC of 0.88.In the training set,RF and SVM outperformed LR,while LR surpassed NB(most comparisons P＜0.05);RF marginally outperformed SVM(P＞0.05).In the testing set,only the fusion model showed significantly higher performance than NB(P＜0.05).DCA revealed that RF and LR multiparametric models yielded the highest net clinical benefit,with excellent calibration.Conclusion Multiparametric DCE-MRI radiomics can effectively differentiate high-and low-grade IDC.The S3-phase radiomic features and multi-model fusion strategy demonstrated superior predictive performance,highlighting their potential for clinical application.