Nomogram based on DCE-MRI radiomics combined with clinical-radiological features in predicting hormone receptor status in breast cancer with low Her-2 expression

Acta Universitatis Medicinalis Anhui 2025, 09, v.60 1745-1754     font:big middle small

Found programs: National Natural Science Foundation of China ( No.82371928) ; Scientific Research Project of Anhui Medical University ( No.2021xkj134) ; Research Project of Anhui Provincial Institute of Translational Medi- cine ( No.2023zhyx-C37)

Authors:Hou Weishu1 ,Pan Hongli1 ,Wang Qun1 ,Li Xiaohu1 ,Yan Yunwen2 ,Yu Yongqiang1

Keywords:breast cancer; dynamic enhanced magnetic resonance imaging; radiomics; human epidermal growth factor receptor 2; hormone receptor; nomogram;

DOI:10.19405/j.cnki.issn1000-1492.2025.09.024

〔Abstract〕 To explore the value of nomogram based on DCE-MRI radiomics combined with clinical-ra- diological features in predicting HR status in breast cancer with Her-2 low expression.Methods A total of 198 pa- tients of Her-2 low expression breast cancer who underwent standardized breast MRI in our hospital from January 2019 to February 2025 were retrospectively analyzed.Patients were divided into HR ( + ) group ( n = 137) and HR ( -) group ( n = 61) .The cases were divided into a training set ( 138 cases) and a testing set ( 60 cases) in a 7 ∶ 3 ratio.Clinical-radiological model was based on clinical and traditional radiological features,radiomics model was based on DCE-MRI,and combined model was constructed,respectively.The nomogram was drawn,and re- ceiver operating characteristic curve was used to compare the performance of different models in predicting HR sta- tus.Results The DCE-MRI radiomics score ( Radscore) between the HR ( + ) group and the HR ( -) group showed statistical differences in both the training and testing sets ( both P<0. 001) .The AUC of the clinical-radio- logical model based on lesion mobility,Ki67,TIC type,enhancement pattern and maximum diameter for predicting HR status in the training set and testing set were 0. 643 and 0. 616,respectively.The AUC of the DEC-MRI ra- diomics model in the training set and testing set were 0. 897 and 0. 860,respectively.The nomogram drawn by combining clinical-radiological features and Radscore showed better predictive performance in both the training set ( AUC = 0. 913) and testing set ( AUC = 0. 898) than the clinical-radiological model ( all P<0. 05) .Conclusion The nomogram combined by DCE-MRI radiomics and clinical-radiological features can effectively predict HR sta- tus of breast cancer with low Her-2 expression,which is helpful to the building of individualized treatment plan for breast cancer patients.