Abstract
BACKGROUND: Accurate assessment of human epidermal growth factor receptor 2 (HER2) status, particularly HER2-low (formerly HER2-negative), is critical for guiding optimal HER2-targeted therapeutic decisions, as these patients may now be eligible for novel anti-HER2 antibody-drug conjugates. This study aimed to develop a radiomic nomogram based on multiparametric magnetic resonance imaging (MRI)-derived radiomic features combined with clinical characteristics for distinguishing HER2-positive and HER2-low breast cancer (BC) from HER2-negative BC (Task 1) and HER2-low from HER2-negative BC (Task 2). METHODS: A total of 364 patients from two centers with invasive ductal carcinoma of BC were retrospectively enrolled from September 2022 to March 2024 and divided into two tasks. Patients from Center 1 (The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University) were randomly assigned to training cohort (Task 1: n=165; Task 2: n=112) and internal validation cohort (Task 1: n=71; Task 2: n=48). Patients from Center 2 (Ganzhou Cancer Hospital) were used as an external validation cohort (Task 1: n=78; Task 2: n=52). Radiomics signatures (RS) models were established using features from dynamic contrast-enhanced (DCE), T2-weighted image (T2WI), and combination (RS-Com). A clinical characteristic model was established through univariate and multivariate analyses, and a radiomics nomogram was developed by integrating radiomics score (Rad-score) with clinically significant characteristics. Its performance was evaluated through metrics such as the area under the curve (AUC), calibration assessment, and decision curve analysis (DCA). RESULTS: For Task 1, RS-Com yielded a greater AUC for training, internal, and external validation of (0.861, 0.784, and 0.794, respectively) than did RS-DCE (AUC =0.743, 0.732, and 0.629, respectively) and RS-T2WI (AUC =0.741, 0.663, and 0.652 respectively). For Task 2, compared with RS-DCE (AUC =0.774/0.668/0.738) and RS-T2WI (AUC =0.771/0.677/0.637), RS-Com also exhibited greater AUCs for training, internal, and external validation (0.822/0.725/0.773). Univariate and multivariate analyses showed that the estrogen receptor (ER) and progesterone receptor (PR) statuses were independent predictors for distinguishing HER2 status. For both Tasks 1 and 2, the radiomic nomogram demonstrated the best performance with AUCs of 0.916/0.940/0.820 and 0.863/0.892/0.833, respectively. CONCLUSIONS: The multiparametric MRI-based radiomic nomogram can more accurately categorize the levels of HER2 expression in invasive ductal carcinoma patients, especially for those with HER2-low expression, serving as an early-stage aid for clinicians to devise tailored and precise therapeutic strategies.