Objective  To explore the diagnostic value of intratumoral area (ITA) and peritumoral area (PTA) ultrasound image-based bioinformatics models for small breast cancer.

Methods  We retrospectively analyzed data of 305 breast lesions from 292 patients with small breast nodules (diameter ≤2 cm) who were treated at People's Hospital of Xinjiang Uygur Autonomous Region between January 2021 and January 2025. The lesions were randomly assigned into the training (214 lesions) and validation sets (91 lesions) in a 7:3 ratio. Radiomics features were extracted from the intertumoral area (ITA) and peritumoral area (PTA) regions at 2, 4, 6, and 8 mm, followed by feature selection and dimensionality reduction. A Logistic regression (LR) algorithm was used to construct a model. The performance of the models were evaluated via receiver operating characteristic (ROC) curve analysis, Hosmer-Lemeshow test, and decision curve analysis (DCA).

Results In the training set, the areas under the ROC curves (AUC) for the ITA, 2 mm PTA, and 2 mm fusion models were 0.869, 0.897, and 0.909, respectively. In the test set, these respective AUC values were 0.813, 0.825, and 0.840. For breast lesions ≤2 cm, <1 cm, and 1-2 cm, the overall accuracies of the 2 mm fusion model were 81.0%, 82.7%, and 80.1%, respectively, whereas the respective overall accuracies of BI-RADS were 76.4%, 81.7%, and 73.6%.

Conclusions ITA and PTA ultrasound imaging-based radiomics models had a high diagnostic value for small breast cancers. The fusion model can effectively improve predictive performance, outperforming the BI-RADS classification in diagnosing small breast lesions of different diameters. Thus, these models have the potential to serve as an auxiliary diagnostic tool in clinical practice.