Objective To construct a risk model based on ultrasound image features and serological indexes for predicting benign and malignant non-mass breast lesions(NML),and to explore its clinical application value.Methods Seventy-eight patients(78 lesions)with pathologically confirmed NML in our hospital were selected,including 36 malignant cases and 42 benign cases.The ultrasound image features including the maximum diameter of the lesion,growth orientation,blood flow signal,architectural distortion,posterior echo attenuation,and microcalcification were obtained by two-dimensional ultrasound and color Doppler ultrasound.The carcinoembryonic antigen(CEA),CA15-3 and miR-194 were obtained by laboratory tests,and their clinical data were retrieved.The differences in ultrasound image features,serological indexes and clinical data were compared between benign and malignant NML.Multivariate Logistic regression analysis was used to screen the independent influencing factors for predicting the benign and malignant NML.A risk model was constructed based on the above independent influencing factors.Receiver operating characteristic(ROC)curve and calibration curve were drawn to evaluate the diagnostic efficacy and calibration degree of the model,respectively.Results The proportions of age≥50 years,lactation history,family history of breast cancer,CEA>2.1 ng/ml,CA15-3>18.7 U/ml,miR-194>1.60,and architectural distortion,blood flow signal grade Ⅱ～Ⅲ,posterior echo attenuation,microcalcification of malignant NML were higher than those of benign NML,and the differences were statistically significant(all P<0.05).There were no statistically significant differences in maximum diameter and growth orientation of the lesions between benign and malignant NML.Multivariate Logistic regression analysis showed that the family history of breast cancer,CEA,CA15-3,miR-194 and architectural distortion,blood flow signal,posterior echo attenuation were independent influencing factors for predicting the benign and malignant NML(all P<0.05).A risk model was constructed based on the above influencing factors.ROC curve analysis showed that the area under the curve for predicting benign and malignant NML was 0.846(95％confidence interval:0.762～0.930),the sensitivity and specificity were 73.50％and 77.30％,respectively.Calibration curve showed that the predicted probability of the model was basically consistent with the actual probability(χ2=8.192,P=0.455),the Brier score was 0.13.Conclusion The risk model based on ultrasound image features and serological indexes has good clinical application value in predicting the benign and malignant NML.
UltrasonographySerological indexNon-mass breast lesions,benign and malignantPredictive model
NETL
NSTL
万方数据
