Objective To explore the sonographic features of subcentimeter breast masses

and to assess the application value by the multi-factor binary Logistic regression.Methods The sonographic features of 497 breast masses confirmed by surgical pathology were analyzed retrospectively.The breast masses were divided into benign group（n=466） and malignant group（n=31） according to the results of pathology.The differences of sonographic features were compared.The multifactor binary Logistic regression analysis was used to screen out the independent influencing factors for the differential diagnosis of benign and malignant subcentimeter breast masses.The regression equation was established and the ROC curve was depicted to analyze the predictive value of Logistic regression model.Results There were statistically differences in morphology

aspect ratio

edge

echo type

in-tumor calcification

peripheral tissue signs and blood flow signal in tumor between the two groups（all P<0.05）

while there was no statistically difference in posterior echo between the two groups（P=0.26）.The multi-factor binary Logistic regression analysis showed that aspect ratio≥1

internal blood flow signal and marginal blood flow signal were independent influencing factors for differentiating benign and malignant breast masses（OR=9.56

9.68

4.29

P=0.02

0.00

0.04）.Logistic regression equation was Logistic（P）=-3.86+2.23×aspect ratio≥1+2.29×internal blood flow signal+1.46×marginal blood flow signal.The Logistic regression model demonstrated that with prediction probablity P=0.05 as cut-off value

the diagnostic accuracy was 95.2%

sensitivity was 83.9%

specificity was 89.1%

and area under ROC curve was 0.89.Conclusion The Logistic regression model based on aspect ratio and blood flow signal can efficiently differentiate malignant subcentimeter breast mass from benign one.