Objective:

To explore the value of logistic regression model based on subjectsclinical information in discriminating breast malignant lesions from benign lesions of Breast Imaging Reporting and Data System (BI-RADS) 4.

Methods:

Retrospectively 221 subjects (133 benign and 88 malignant) confirmed by histopathology were recruited whose BI-RADS grade was 4 and thecl

inical information were collected. Logistic regression analysis was used to screen the clinical information features that can discriminate malignant from benign lesions and a regression model was established. The comparison was made between regression model combined with BI-RADS and BI-RADS classification alone for differential diagnosis between malignant and benign lesions.

Results:

Nine clinical information features were found to be related to malignant and benign lesions. Three features of whether the lesion can be touched (OR=7.196)

whether the lesion position was fixed (OR=10.150)

and whether the maximum diameter of the lesion was more than 2 cm (OR=4.208) have a higher risk than other clinical information (

P

＜0.05). Using BI-RADS classification alone

the diagnostic sensitivity

specificity and accuracy were 86.3%

69.9% and 76.5%; the diagnostic sensitivity

specificity and accuracy of regression model combined with BI-RADS were 88.6%

73.7% and 79.6%.

Conclusion:

Logistic regression model based on subjectsclinical information combined with BI-RADS classification is helpful to improve the diagnostic efficiency of malignant and benign lesions and further to reduce unnecessary benign biopsy.