The vlue ofintratumoral and peritumoral radiomics of dynamic contrast-enhanced MRI in predicting the vascular inva-sion status of invasive ductal carcinoma of the breast

XIE Hanmin; HUANG Huang; TANG Dawei; TAN Yu; HUANG Yubin

doi:10.20258/j.cnki.1006-9011.2025.11.015

Abstract：Objective To explore the application value of radiomics models based on the intratumoral and peritumoral re-gions of dynamic contrast-enhanced magnetic resonance imaging(DCE-MRI)in predicting the lymphovascular invasion(LVI)status of invasive ductal carcinoma(IDC)of the breast.Methods A total of 219 patients with pathologically confirmed breast IDC were retrospectively included in this study.Their imaging and clinical data were collected and randomly divided into a train-ing set(175 cases)and a validation set(44 cases)at a ratio of 8∶2.Based on DCE-MRI images,the regions of interest(ROIs)for the intratumoral region were semi-automatically delineated layer by layer,and the peritumoral region was defined by automati-cally extending the intratumoral ROI outward by 4 mm.Radiomics features of the intratumoral and peritumoral regions were ex-tracted using the Pyradiomics software library.The optimal features were selected using t-tests,minimum Redundancy Maximum Relevance(mRMR),and Least Absolute Shrinkage and Selection Operator(LASSO)regression,and were incorporated into the ExtraTrees machine learning classifier to construct radiomics prediction models based on the intratumoral,peritumoral,and com-bined intratumoral and peritumoral regions.The predictive performance of each model was evaluated using the Receiver Operat-ing Characteristic(ROC)curve,and the differences between the models were compared using the Delong test.Results In the training set,only the maximum tumor diameter was statistically different between the LVI positive group and the LVI negative group(P<0.05).There were no significant differences in age,estrogen receptor(ER),progesterone receptor(PR),human epi-dermal growth factor receptor 2(HER-2),Ki-67,gland type,tumor shape,tumor margin,spiculation,axillary lymph node me-tastasis,time-signal intensity(TIC)curve type and molecular typing between the two groups.(all P>0.05).The Area Under the Curve(AUC)values of the intratumoral model in the training set and validation set were 0.839 and 0.733,respectively,while,those of the peritumoral model were 0.788 and 0.622,respectively.The combined intratumoral and peritumoral model had the best predictive performance,with AUC values of 0.840 and 0.771 in the training set and validation set,respectively.Conclu-sion The radiomics models based on DCE-MRI intratumoral and peritumoral regions have good application value in predicting the LVI status of breast IDC.