Preliminary study on DCE-MRI radiomics analysis for differentiation of HER2-low and HER2-zero breast cancer

Abstract Purpose This study aims to evaluate the usefulness of radiomic features obtained by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in distinguishing HER2-low from HER2-zero breast cancer. Patients and methods: We performed a retrospective analysis of 118 MRI cases, including 78 HER2-low and 40 HER2-zero patients confirmed by immunohistochemistry or fluorescence in situ hybridization. For each case in the DCE-MRI phase, a region of interest (ROI) was determined and 960 radiomics were extracted. Lasso regression was used to identify similar features with HER2-low and HER2-zero variants. The effectiveness of the model in distinguishing between HER2-low and HER2-zero was assessed using logistic regression (LR). Additionally, an integrated radiological model was developed to include Rad scores obtained from DCE-MRI and clinic-radiological semantic features and visualized as a radiomics nomogram using logistic regression model. Results The logistic regression model demonstrated excellent performance, attaining area under the curve (AUC) values of 0.875 and 0.845 on the training and testing sets, respectively, outperforming the clinical model on both sets (AUC = 0.691 and AUC = 0.672). Higher HER2 risk factors were associated with increased Rad-score and Time intensity curve (TIC). In both sets, the radiomics nomogram performed better than models containing only clinic-radiological semantic features or radiomics signatures, with AUC, sensitivity, and specificity values ​​of 0.892 and 79.6% and 82.8% in the training set, and 0.886, 83.3%, and 90.9% in the testing set respectively. Conclusions The combined radiomic nomogram based on DCE-MRI demonstrated promising potential in predicting the difference between HER2-low and HER2-zero status in breast cancer patients.