Objective To explore the CT-based radiomics in the differential diagnosis of spinal osteosarcoma and chon-drosarcoma.Methods 10 patients with spinal osteosarcoma and 28 patients with chondrosarcoma who were admitted were retrospectively analyzed.Image segmentation and feature extraction were performed on CT image with the Research Portal V1.1.Correlation coefficient and the least absolute shrinkage and selection operator(LASSO)algorithm were used to select the radiomics features,a fivefold cross-validation was repeated 20 times with random initialization to reduce bias.The receiv-er operating characteristic curve(ROC)was used to evaluate the performance of the models.Results The 2264 ra-diomics features were selected,and the most significant 7 radiomics features were obtained to construct radiomics model.The logistic model based on CT can yield the the mean area under the ROC curve(AUC)of 0.928(95％CI 0.806-1)in training cohort,and 0.912(95％CI 0.802-1)in test cohort.Conclusion CT-based radiomics models have some poten-tial for differentiating spinal osteosarcoma from chondrosarcoma.
NETL
NSTL
万方数据
