Fig. 2From: Exploring non-invasive precision treatment in non-small cell lung cancer patients through deep learning radiomics across imaging features and molecular phenotypesComparison of receiver operating characteristic (ROC) curves for three models predicting prognostic risk and gene mutations, including LVI (a), PI (b), and T staging (c), as well as for EGFR (d), KRAS (e), ALK (f), TP53 (g), PIK3CA (h), and ROS1 (i). The corresponding area under the curve (AUC) values are indicated in parentheses next to the ROC curvesBack to article page