Prediction of HER2-Low Breast Cancer via Multimodal Ultrasound Imaging

Human epidermal growth factor receptor 2 (HER2) is a key biomarker for clinical management and prognostic evaluation of breast cancer patients. This study was aimed at assisting the preoperative and non-invasive prediction of HER2-low breast cancer using multimodal ultrasound imaging and clinicopathological indicators, providing valuable imaging information for clinical precision diagnosis and personalized treatment strategies, especially in the application of antibody-drug conjugates such as T-DXd.

This retrospective study included 147 pathologically confirmed breast cancer patients from two institutions: 101 in the training set and 46 in the external validation set. All patients underwent multimodal ultrasound (grayscale, color Doppler, elastography, and contrast-enhanced imaging) and had complete clinicopathological data. Patients were categorized as HER2-negative, HER2-low, or HER2-positive based on immunohistochemistry. Logistic regression was used to construct predictive models.

Compared with the HER2-negative group, low Ki-67, PR positivity, longer rise time (RT), and lower Emax values were independent predictors of HER2-low status (p < 0.05), yielding an AUC of 0.876, sensitivity 0.833, and specificity 0.781. Compared with HER2-positive cancers, HER2-low cases showed low Ki-67, ER/PR positivity, low Emax, and a DVPC pattern characterized by an initial increase followed by a subsequent decline as independent predictors (p < 0.05), with an AUC of 0.929, sensitivity 0.905, and specificity 0.856. External validation confirmed robust model performance (AUC = 0.925 and 0.918 for HER2-low vs. negative and positive, respectively).

A model integrating multimodal ultrasound and clinicopathological factors effectively predicts HER2-low breast cancer, offering valuable imaging-based support for clinical decision-making.