| Purpose:Breast cancer is the most common type of cancer in the worldwide,and its incidence has been increasing in recent years,with the second highest mortality rate after lung cancer and is a serious therat to women’s lives and health worldwide.The status of axillary lymph nodes in breast cancer affects patient treatment decisions and prognosis.Accurate and non-invasive preoperative assessment of axillary lymph node metastasis status has always been a challeng for clinicians.Axillary lymph node dissection(ALND)and sentinel lymph node biopsy(SLNB)are considered to be the standard methods to assess axillary lymph node status,but both methods are invasive and postoperative complications should not be underestimated,such as upper limb lymphedema,shoulder numbness and many other complications that seriously affect patients’ quality of life.The extraction of image information through radiomics to assist in disease diagnosis and treatment decisions has become a focus of research in the field of scientific research.Digital breast tomography(DBT)is a common imaging examination mothod for breast cancer,and its radiomics features may correlate with axillary lymph node status.The aim of this study was to retrospectively analyze DBT images of breast cancer patients,select features that are highly correlated with axillary lymph nodes in breast cancer from the many features extracted,and use these features to develop a breast DBT-based prediction model for preoperative prediction of ALN metastasis status in breast cancer patients,providing an effective method for clinicians to accurately assess axillary lymph node status preoperatively.Materials and methods:1.Retrospectively collected 120 breast cancer patients who visited The third bethunt hosopital of Jili university for breast lumps and underwent surgery from December 2019 to August 2022,all patients underwent DBT examinaton before surgery and have complete basic clinical data,where basic clinical data were obtained from electronic medical records or pathology data,including axillary lymph node metastasis status,age,family history,lesion location,size,ER status,PR status,HER-2 status,and Ki67 proliferation index.All the above patients were randomly divided into training and validation sets in a ratio of 7:3.They were divided into two categories according to the presence or absence of lymph node metastasis,lymph node positive group(49 patients)and negative group(71 patients).2.The original DBT DICOM images of all patients were acquired and the region of interest(ROI)was manually outlined in the DBT images(including the mediolateral oblique(MLO)view and cranial caudal(CC)view)by applying ITK-SNAP software.3.All outlined completed images were used to extract radiomics features from CC and MLO images using the open source software package Pyradiomics,respectively,and to build the radiomics score(Radscore)prediction model after radiomics feature selection.Radscore for each patient was calculated as a linear fit to the selected features,which were weighted by their respective coefficients,so that the radiomics score(Radscore)contained all the information of the selected features.4.Three models,Radscore＿cc,Radscore＿mlo,and Radscore＿combine,were obtained to predict the presence or absence of metastasis in axillary lymph nodes.The diagnostic efficacy of the models was assessed by ROC curves,AUC values,sensitivity,and specificity.Result:In the training set,the accuracies of Radscore＿cc,Radscore＿mlo,and Radscore＿combine were 79.8%、71.4%、83.3%,with AUC values of 0.821,0.780,and 0.890,respectively;in the validation set,the accuracies of Radscore＿cc,Radscore＿mlo,and Radscore＿combine were 63.9%、75.0%、75.0%,with AUC values of 0.750,0.885,and 0.892,respectively.In which the Radscore＿combine model had the highest prediction efficiency,and the AUC values were higher than those of Radscore＿cc and Radscore＿mlo two models.Conclusion:Preoperative breast DBT-based radiomics model is a reliable and non-invasive tool for predicting preoperative breast cancer axillary lymph node metastasis,and has good predictive efficacy for optimizing current treatment strategies for breast cancer patients. |
PDF Full Text Request