Objective To analyze the features of contrast-enhanced ultrasound（CEUS） of benign and malignant breast lesions

construct a predictive model

and to explore the value of this model for risk assessment of breast imaging report and data system（BI-RADS） 4 category of lesions.Methods A total of 174 patients with a total of 180 lesions（BI-RADS 4 category）who underwent breast ultrasound examination in our hospital were selected.All patients obtained the final pathological results through surgery.The cases were divided into model construction group（91 cases with 94 lesions） and model verification group（83 cases with 86 lesions）.Single-factor and multi-factor Logistic regression analysis was performed on the characteristics of the CEUS enhancement mode of the model building group to construct a predictive model and draw the ROC curve.The diagnostic efficacy of the predictive model for benign and malignant breast lesions was calculated

and the consistency of the blind reading results was analyzed by ultrasound doctor with different experience.Results The characteristic risk factors of Logistic regression analysis were "crab foot" sign or vascular distortion sign

enhanced uniformity and enhanced focal area（OR=11.308

5.980

3.377

all P

<

0.01）.The Logistic regression equation（the predictive model）:y=-4.239+2.425 x

8

+1.788 x

3

+1.217 x

5

（x

8

: "crab foot" sign or vascular distortion sign

x

3

:enhanced uniformity

x

5

:enhanced focal area）.The area under ROC curve of the CEUS predictive model in distinguishing benign and malignant breast lesions was 0.909

the sensitivity

specificity

accuracy

positive predictive value and negative predictive value of the CEUS predictive model were 95.83%

81.58%

89.53%

86.79% and 93.94%

respectively.The Kappa value of the blind reading result consistency test between the different ultrasound doctor was 0.683.Conclusion Breast CEUS prediction model has a high risk assessment value for BI-RADS 4 category of breast lesions

and different ultrasound doctor use this prediction model to diagnose benign and malignant breast lesions has high consistency.