Figure 3.

Radiomics features selection. (A) The tuning parameter (λ) selected in the LASSO algorithm through 10-fold cross-validation. The binomial deviance was drawn versus log (λ). Two vertical lines represent the optimum values using the minimum criteria and 1 standard error of the minimum criteria (1-SE criteria). The optimal λ value of 0.067 was picked. (B) LASSO coefficient profiles of 116 radiomics features. A vertical line was drawn at the value selected using 10-fold cross-validation, where optimal λ generated four coefficients.