OBJECTIVES: To investigate the value of unenhanced CT-based radiomics in determining whether bone metastases have progressed in breast cancer patients.

METHODS: After 2-3 cycles of drug treatment, unenhanced CT was performed in 134 patients with bone metastasis of breast cancer from 3 different hospitals. According to whether the disease progressed or not, the images were divided into invalid group and valid group. The maximum cross-section of the largest osteolytic lesions in CT images was used as the region of interest (ROI) for feature extraction. Variance threshold, SelectKBest and least absolute shrinkage and selection operator (LASSO) were used to reduce the dimension of features. On this basis, the K- nearest neighbor algorithm (KNN), support vector machine (SVM), eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Logistic Regression (LR) and Decision Tree (DT) were trained to establish radiomics models. Finally, the receiver operating characteristic (ROC) curves were generated to determine the diagnostic performance of various models.

RESULTS: The KNN classifier demonstrated a more optimal performance compared to other models The areas under the ROC curve (AUC) in validation set was 0.810.

CONCLUSIONS: The radiomics provides a new idea for evaluating the efficacy of patients with bone metastases, KNN model has the best performance.