Fig. 4.

Radiomics feature selection using the LASSO regression algorithm in the primary cohort. A. Selection of the parameter (λ) in the LASSO model via 10-fold cross-validation depending on the minimum criteria. The binomial deviance curve versus log (lambda) was plotted, and the left vertical line corresponds to the optimal value of the minimum criterion; the right vertical line corresponds to the optimal value of the 1-SE criteria. The optimal λ value of 0.0127with threshold log (λ) of was − 4.32 was selected. B. LASSO coefficient profiles of the 55 features. Vertical line was plotted at the value selected using 10-fold cross-validation, where optimal λ resulted in 14 nonzero coefficients. C. The receiver operating characteristic curves (ROC) of the radiomics signature in the training and validation cohorts