Abstract Objective： To compare the diagnostic value of predictive models based on two-dimensional ultrasound texture features (2D-Ultrasomics) and contrast-enhanced ultrasound (CEUS) for distinguishing benign and malignant breast nodules ≤2 cm in diameter. Methods： A total of 109 patients (112 nodules) with breast nodules who underwent 2D-CDUS and CEUS at our hospital were included. Based on pathological results, the patients were divided into benign (58 nodules) and malignant (54 nodules) groups. The 2D-CDUS and CEUS image features were compared between the groups. 2D-Ultrasomics features were extracted from 2D ultrasound images, followed by feature selection using the Least Absolute Shrinkage and Selection Operator (LASSO). Multivariable logistic regression was used to construct three models: 2D-CDUS, 2D-CDUS combined with CEUS (2D-CD+CEUS), and 2D-CDUS combined with ultrasound texture features (2D-CDUS+Ultrasomics). Receiver Operating Characteristic (ROC) curves were drawn to assess the diagnostic efficacy of each model for distinguishing benign and malignant breast nodules ≤2 cm. The Hosmer-Lemeshow test was used to evaluate model fit, and clinical decision curve analysis was performed to assess the clinical applicability of the models. Results： The comparison of 2D-CDUS (echogenicity, boundary, blood flow, width) and CEUS (enhancement pattern, phase, boundary, uniformity, and lesion range) image features between the two groups showed statistically significant differences (all P ＜0.05). A total of 818 2D-Ultrasomics features were extracted, with 6 retained after LASSO selection. Three models were constructed: the 2D-CDUS model (based on boundary and width), the 2D-CDUS+CEUS model (based on boundary, width, and contrast boundary), and the 2D-CDUS+Ultrasomics model (based on width, gray-level run length matrix, gray-level dependence matrix, and gray-level size zone matrix). ROC analysis revealed the 2D-CDUS+Ultrasomics model had the highest AUC (0.917) for distinguishing benign and malignant nodules ≤2 cm, significantly higher than the 2D-CD+CEUS model (0.892) and the 2D-CDUS model (0.823) (all P < 0.001). The Hosmer-Lemeshow test showed a good model fit for all three models (P = 0.818, 0.103, 0.281). Clinical decision curve analysis showed the 2D-CDUS+Ultrasomics model provided the highest clinical benefit in the thresholds of 0.20–0.39, 0.43–0.78, and 0.88–0.91. Conclusion： The 2D-Ultrasomics-based predictive model can more accurately distinguish between benign and malignant breast nodules ≤2 cm in diameter compared to the CEUS-based predictive model, thereby aiding in the early and accurate diagnosis of small breast nodules and facilitating treatment decision-making. Keywords: Ultrasound examination; Contrast agent; Texture features; Breast nodules; Benign or malignant