Abstract: Objective To investigate the value of models based on preoperative ultrasound characteristics, inflammatory indicators and ultrasound radiomics features in predicting axillary lymph node(ALN)metastasis of breast cancer. Methods The breast ultrasound images and clinical data of 175 breast cancer patients were retrospectively analyzed. The 3D Slicer software was used to outline the region of interest and extract the radiomics features. The interclass correlation coefficient, Pearson correlation coefficients and recursive feature elimination were used to select the features. After the radiomics score(Radscore)was calculated, the radiomics model was constructed. The clinical model was constructed by screening clinical risk factors through univariate and multivariate Logistic regression, and then the Radscore was added to construct a combined prediction model. The predictive efficacy and clinical value of the models were assessed with the receiver operating characteristic(ROC)curve, calibration curve, and decision curve. Results Eighteen radiomics features were included in the radiomics model, and tumor size, ultrasound ALN status and platelet to lymphocyte ratio(PLR)were included in the clinical model. The tumor size, ultrasound ALN status, PLR and Radscore were included in the combined prediction model. The combined prediction model had the highest prediction efficacy. In the training and validation sets, the area under the curve(AUC)of the combined prediction model were 0.935 and 0.858, respectively. Conclusion The combined prediction model based on tumor size, ultrasound ALN status, inflammatory indicator PLR and ultrasound radiomics is effective in predicting ALN metastasis in breast cancer patients.

Key words: Breast cancer, Axillary lymph node metastasis, Ultrasonography, Radiomics, Inflammatory indicators