Objective：To compare the diagnostic value of CEUS 5-point method and CEUS predictive models for breast imaging report and data system（BI-RADS） 4 of breast lesions with different sizes. Methods：165 patients (166 lesions in total) with breast lesions confirmed by puncture biopsy or surgical pathology in our hospital were collected.All patients were examined by conventional ultrasound and CEUS, and BI-RADS was classified into 4 categories.According to the size of the lesions, 101 lesions were classified into the group with diameter ≤ 2.0cm and 65 lesions were classified into the group with diameter > 2.0cm.The CEUS characteristics of the lesions were analyzed,including enhanced intensity，enhanced phase，enhanced mode，enhanced morphology，enhanced boundary，enhanced homogeneity，perfusion defect，enhanced range，crab foot sign，and nourishing vessels.The CEUS 5-point method and CEUS predictive models were used to re-determine the benign and malignant lesions, and receiver operating characteristic (ROC) curve was drawn to compare the diagnostic efficacy of the two methods for the benign and malignant of BI-RADS 4 breast lesions with different sizes. Results：The 10 CEUS characteristics of benign and malignant lesions with diameter ≤ 2.0cm were statistically significant differences（all P<0.05）.CEUS characteristics of benign and malignant lesions with diameter > 2.0cm，except for enhanced intensity，enhanced mode and perfusion defect，there were statistically significant differences in the 7 enhancement characteristics（all P<0.05）.ROC curve analysis showed that for lesions with diameter ≤ 2.0cm，the area under the curve（AUC） of CEUS 5-point method and CEUS predictive models for diagnosis of benign and malignant lesions were 0.884 and 0.861，respectively，with no statistical significance.For lesions with diameter > 2.0cm，the AUC of benign and malignant lesions diagnosed by CEUS 5-point method and CEUS predictive models were 0.932 and 0.864，respectively，with statistically significant difference（P<0.01）.The diagnostic coincidence rate of lesions with 2，3 and 5 scores in CEUS 5-point method was high，while the diagnostic coincidence rate of lesions with 1 and 4 scores was low；Among CEUS predictive models，A，D，E and F models had better predictive performance，followed by B model and C model. Conclusions：For BI-RADS 4 breast lesions with diameter ≤ 2.0cm，the CEUS 5-point method is equivalent to the diagnostic value of the CEUS predictive models；for BI-RADS 4 breast lesions with diameter > 2.0cm，the CEUS 5-point method has higher diagnostic value.