A machine learning-based radiomics model for the prediction of axillary lymph-node metastasis in breast cancer

Objective: The aim of this study was to develop and validate machine learning-based radiomics model for predicting axillary lymph-node (ALN) metastasis in invasive ductal breast cancer (IDC) using F-18 fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT).

Methods: A total of 100 consecutive IDC patients who underwent surgical resection of primary tumor with sentinel lymph-node biopsy and/or ALN dissection without any neoadjuvant treatment were analyzed. Volume of interests (VOIs) were drawn more than 2.5 of standardized uptake value in the primary tumor on the PET scan using 3D slicer. Pyradiomics package was used for the extraction of texture features in python. The radiomics prediction model for ALN metastasis was developed in 75 patients of the training cohort and validated in 25 patients of the test cohort. XGBoost algorithm was utilized to select features and build radiomics model. The sensitivity, specificity, and accuracy of the predictive model were calculated.

Results: ALN metastasis was found in 43 patients (43%). The sensitivity, specificity, and accuracy of F-18 FDG PET/CT for the diagnosis of ALN metastasis in the entire patients were 55.8%, 93%, and 77%, respectively. The radiomics model for the prediction of ALN metastasis was successfully developed. The sensitivity, specificity, and accuracy of the radiomics model for the prediction of ALN metastasis in the test cohorts were 90.9%, 71.4%, and 80%, respectively.

Conclusion: The machine learning-based radiomics model showed good sensitivity for the prediction of ALN metastasis and could assist the preoperative individualized prediction of ALN status in patients with IDC.