Fig. 3

The process of the ultrasound radiomic feature selection. (A) Flow chart of ultrasound radiomic features selection. (B) Radiomic feature selection by applying the LASSO binary logistic regression model. The vertical lines on the left side of the horizontal coordinate represent the tuning parameter lambda (λ) chosen in the LASSO model using 10-fold cross-validation via minimum criteria. (C) Coefficient convergence plot of the screened features, with the best log (λ) value screening out 12 histological features with non-zero coefficients. LASSO, least absolute shrinkage, and selection operator; PTC, papillary thyroid carcinoma; AIC, Akaike information criterion