ObjectiveTo explore the application value of mammography radiomics nomogram in predicting the expression status of breast cancer Her-2 before surgery.

MethodsA retrospective analysis was performed for 262 women with invasive ductal carcinoma (IDC) who underwent a mammogram before surgery or puncture.According to the 7∶3 ratio, 183 cases were randomly divided into training set, and 79 cases into test set.Region of interest (ROI) was manually delineated by mammogram image, radiomics features were extracted by least absolute shrinkage and selection operator (LASSO) regression, dimensionality reduction was retained through statistical and LASSO machine learning methods, and logistic regression was used as a classifier to estabish radiomics models; combined with image data, single-multivariate logistic regression was used to screen independent risk factors to establish image feature models.The radiomics nomogram model was established by combining the radiomics features with independent risk factors.The ROC curve was used to analyze the predictive performance of each model, calculate the area under the curve (AUC), and draw the calibration curve and decision curve to evaluate its efficiency.

ResultsThe nomogram model had the best prediction performance, with the training set sensitivity of 84.62% and specificity of 84.75%, the AUC value of 0.920, the sensitivity of the test set of 84.00%, the specificity of 83.33%, and the AUC value of 0.916.In the calibration curve, the prediction curve of the nomogram model was in good agreement with the ideal curve, and the decision curve had a good net benefit.

ConclusionsMammograms can be a useful tool for preoperative assessment of Her-2 status in breast cancer patients.