Objective：To develop a predictive model based on clinical and ultrasound characteristics for differentiating between benign from malignant small breast masses with non-parallel growth and classified by the Breast Imaging Reporting and Data System（BI -RADS）as BI-RADS 4A and 4B，and to evaluate its application value. Methods：For this retrospective study，327 patients with breast masses were recruited in the First Affiliated Hospital of Nanjing Medical University from June 2020 to May 2024. These patients were screened and diagnosed with BI-RADS 4A and 4B classifications by ultrasound，with non-parallel growth and a diameter of ≤10 mm. They were randomly divided into a training set（n=229）and a validation set（n=98）at the ratio of 7∶3. Logistic regression analysis was used to identify risk factors for differentiating benign from malignant breast lesions and to construct a diagnostic predictive model. The effectiveness of the model was evaluated by the receiver operating characteristic（ROC）curve and the decision curve analysis（DCA）. Results：Among the 327 cases，36.1%（118/327）were malignant. Univariate and multivariate logistic regression analyses identified age，margin，elasticity assessment，and US - BI - RADS classification as independent risk factors for malignant masses. A diagnostic model incorporating these four variables was established and presented as a nomogram. The area under curve（AUC）of the ROC was 0.846 and 0.798 for the training and validation sets，respectively. DCA confirmed that the model’s prediction of the risk of benign and malignant masses could benefit patients. Moreover，risk stratification based on nomogram-calculated risk scores for all patients showed that those with risk scores ≥0.7（top 20%）were classified as high-risk，while those with scores ≤0.1（bottom 20%）were classified as low - risk，with malignant rates of 8.8% and 82.1%，respectively. Conclusion：The predictive model constructed based on clinical and ultrasound characteristics effectively differentiates benign from malignant small breast masses of BI-RADS 4A and 4B categories with non-parallel growth. Risk stratification based on nomogram-calculated risk scores indicates a malignant rate of only 8.8% in the low- risk group and 82.1% in the high-risk group，demonstrating significant clinical value.