Figure 2.
PD-L1 expression signature assessment results for high and low expression classification. (A, B and C) were the ROC curves of the training, validation, and test cohorts, respectively. (D, E and F) were the decision curves of the training, validation, and test cohorts, respectively. Among them, each figure contained the results of PD-L1 expression signature (PD-L1ES), deep learning model (DLM), radiomic model (RM) and clinical model (CM). They all showed that PD-L1 expression signature was superior to other models.