Ultrasound Habitat Radiomics Plus Clinical Features for Pre‐Operative Prediction of Sentinel Lymph Node Metastasis in Breast Cancer

Objectives This study aimed to develop a multimodal model that integrates habitat radiomics and clinical features for non‐invasive prediction of axillary sentinel lymph node (SLN) metastasis in breast cancer (BC). Methods We retrospectively analyzed ultrasound images and clinical data from 191 female patients with BC treated at Beijing Luhe Hospital, Capital Medical University from May 2023 to January 2025. Patients were randomly assigned to training and test sets at a 7:3 ratio. Four models were constructed: a traditional radiomics model, a clinical model, a habitat model characterizing the tumor microenvironment, and a multimodal model combining habitat radiomics with clinical features. Model performance was assessed using receiver operating characteristic curve analysis and decision curve analysis. Results In the training set, the habitat model achieved an area under the curve (AUC) of 0.869, outperforming the clinical model (AUC = 0.718) and the traditional radiomics model (AUC = 0.771). The multimodal model, integrating habitat analysis and clinical features, yielded the highest AUC (0.900). In the test set, the habitat model again showed superior discriminative ability (AUC = 0.866) compared with the clinical model (AUC = 0.727) and the radiomics model (AUC = 0.662); meanwhile, the multimodal model achieved the highest AUC (0.902). Conclusion The habitat model demonstrated superior predictive performance compared with models based solely on ultrasound radiomics or clinical features, and the multimodal approach yielded the best overall accuracy. This combined model offers a promising non‐invasive tool for pre‐operative assessment of SLN status in patients with BC and may assist clinical decision‐making.