Fig. 3

Model performance and validation. (A) Receiver operator characteristic (ROC) curve of the radiomics prediction model. Radiomic score distribution in the (B) training cohort and D) testing cohort. The x-axis represents the patient ID, so each column corresponds to one patient. The y-axis represents the probability of a KRAS mutation predicted by the radiomic model. The black columns indicate that the patients actually had KRAS mutations, while the grey columns indicate that the patients had wild-type KRAS. In the prediction model, a higher column indicates a higher probability for a KRAS mutation. Decision curve analysis (DCA) in the (C) training cohort and E) testing cohort. The x-axis represents the risk threshold of the radiomic prediction model, while the y-axis represents the net benefits of each threshold. The solid black line represents the radiomic prediction model.