The Application Of FFDM Radiomics In Predicting Axillary Lymph Node Metastasis Of Breast Cancer

Objective:To explore the value of full-field digital mammography(FFDM)based radiomics in predicting axillary lymph node(ALN)metastasis in breast cancer patients.Methods:The FFDM images and clinicopathological data of primary breast cancer patients in the first hospital of Shanxi Medical University from September 2016 to September 2021 were retrospectively analyzed.A total of 407 patients were included and randomly divided into training set and test set in a ratio of 7:3.Tumor lesions on the FFDM images were segmented manually and radiomics features were extracted.Feature screening was performed by least absolute shrinkage and selection operator(LASSO)regression to build the radiomics model and calculate the radiomics score(Rad-score).Clinicopathological and X-ray features between the ALN non-metastatic group(n=242)and ALN metastatic group(n=165)were analyzed by logistic regression,and the clinical model was developed.The combined prediction model was constructed based on clinical risk factors and Rad-score,and the nomogram was drawn.Calibration curve and receiver operating characteristic(ROC)curve was used to evaluate the performance of the model in the training set and test set,and the clinical utility of the model was assessed by decision curve analysis(DCA).Results:Age of diagnosis,tumor size,menstrual status,lymph node palpation status,Ki67 expression,histological grade,tumor morphology,tumor margin,and lymph node status observed by FFDM differed between the ALN non-metastatic group and ALN metastatic group(P<0.05).Multi-factor logistic regression analysis showed that tumor size(P=0.003),tumor morphology(P=0.025),lymph node palpation status(P=0.002),lymph node status observed by FFDM(P<0.001)were selected to construct the clinical model.Eleven radiomics features were screened to construct the radiomics model.And the nomogram model included tumor size,tumor morphology,lymph node palpation status,lymph node status observed by FFDM and Rad-score.The area under curve(AUC)of clinical model,radiomics model and combined model in the training set were 0.820,0.757 and 0.873,respectively,and the AUC in the test set were 0.809,0.745 and 0.829,respectively.The predictive performance of the combined model is better than that of the single radiomics model and clinical model,also better than radiologists in diagnosing ALN metastasis on FFDM images.Conclusion:The combined model constructed by FFDM-based radiomics features and clinical risk factors can be an important predictive tool to help doctors assess the ALN status of breast cancer patients preoperatively.