Fig. 4

Radiomics feature selection using least absolute shrinkage and selection operator (LASSO) regression in pairwise tissues. The area under the receiver operating characteristic curve (AUC) was plotted by tuning the optimal parameter (λ) selection via 10-fold cross-validation and minimum criteria in the LASSO model. The left and right dotted lines denote the minimum criteria and the 1-standard error criterion (1-SE) in the pairwise tissues, including esophageal squamous cell carcinoma vs. proximal tumor-adjacent (PTA) tissue, the tumor vs. proximal tumor-distant (PTD) tissue, and the PTA vs. PTD tissue (A, B and C, respectively). The LASSO coefficient profiles the 1013, 1050 and 823 radiomics features for the differentiation of previous pairwise tissues, and as a result, 20 19 and 5 nonzero coefficients have been produced (D, E and F, respectively)