Figure 1.

The study process: The most representative image of each tumor on thyroid 2D ultrasound (US) image was selected. Radiomics features including first order, shape, gray-level cooccurrence matrix (GLCM), and gray-level size zone matrix features (GLSZM) were extracted. A radiomics signature was generated using the least absolute shrinkage and selection operator (LASSO) and was used to train a support vector machine (SVM) classifier in five-fold cross-validation. For determining association between the radiomics signature, US findings, clinicopathological variables, and distant metastasis, univariate and multivariate logistic regression analyses were performed. Another SVM classifier was built using significant variables from the multivariate analysis.