The Application Of Logistic Regression Analysis In The Diagnosis Of Benign And Malignant Breast Masses By The Ultrasonic Description Of The Word Listed In BI-RADS-US(2013)

Objective Logistic regression model was used to analyze the clinical data and ultrasonic characteristics which was listed in Breast Imaging Reporting And Data System(BI-RADS-US,promulgated in 2013), to screen out the more important imaging performance and clinical dataMethod 367 patients were examined by gray-scale ultrasound,CDFI, Ultrasound Elastic imaging in the Second Affiliated Hospital of Anhui Medical University. A total of 430 breast masses were found. Combining with BI-RADS-US promulgated in 2013 recorded the ultrasonic characteristics: size, shape, orientation, edge, margin, echo pattern,posterior acoustic features, micro calcifications in mass, gross calcifications in mass, catheter calcification, calcification outside the mass, architectural distortion, duct changes, skin changes, edema, vascularity, elasticity assessment, enlarged axillary lymph nodes. Refer to the patient’s medical records, record the clinical data: patients,family history of cancer and age, the masses, position. Comparison with pathologic results, apply SPSS17.0 statistical software, in the single factor analysis: the Quantitative data were analyzed by independent sample t-test or samples nonparametric tests; the qualitative data were analyzed by chi-square test. When R is less than 0.05, he difference is statistically significant. The variables which had statistical significance were analyzed using binary logistic regression analysis of enter method analysis.Established Prediction model, use Hosmer and Lemeshow test to examine the fitting of the model and evaluated the diagnostic significance.Result1.The results of univariate analysis:malignant lesions were 17, benign lesions were 260.Age. A total of 17 indicators can be used to identify benign and malignant breast masses: shape, orientation, edge,margin, echo pattern, posterior acoustic features, micro calcifications in mass,gross calcifications in mass, catheter calcification, calcification outside the mass, architectural distortion, duct changes, skin changes, edema, vascularity, elasticity assessment,enlarged axillary lymph nodes have statistically significant.2.The results of multivariate analysis:7 main risk factors for breast cancer diagnosis were screened out, ordered according to the relative risk(OR value) from high to low:micro calcifications in mass(OR=14.878), margin spiculated(OR=13.326), margin microlobulated(OR=12.475),margin angular(OR=11.843), age ≧40(OR=4.652), enlarged axillary lymph nodes(OR=4.410), nor parallel(OR=3.424). The Logistic regression model: Logit(P) =-21.311+1.537 ×(age≧40) + 1.231 × nor parallel + 2.472 × margin agular +2.524×margin microlobulated + 2.590 × margin spiculated + 2.700 × micro calcification in mass + 1.484 × enlarged axillary lymph nodes.Conclusions1.The indicators having statistically significant in the result of single factor analysis included all ultrasound signs listed in the BI-RADS-US(2013), that once again proved that BI-RADS-US is helpful for the differentiation of benign and malignant breast masses.2.micro calcifications in mass, margin spiculated, margin microlobulated, margin angular, age ≧40, enlarged axillary lymph nodes, nor parallel were the main risk factors in the diagnosis of breast cancer. The more risk factors, the more likely to be malignant breast masses.3.The establishment of Logistic regression model proves that the combined application of various ultrasonic signs and clinical data can effectively identify benign and malignant breast masses, and has high practical value.