Fig. 4

Development and performance of the radiomics nomogram. (A) Nomogram based on the 2D corrected radiomics features. (B) Calibration curves of the nomograms built on 3D uncorrected, 3D corrected, 2D uncorrected 2D corrected radiomics features in the training cohort. (C) Calibration curves of the nomograms built on 3D uncorrected, 3D corrected, 2D uncorrected 2D corrected radiomics features in the validation cohort. The calibration curves suggesting the perfect match between the actual (Y-axis) and nomogram-predicted (X-axis) responders. (D) Decision curves showed relatively good performance for the models in terms of clinical application and indicated that all the models added more benefit than either the treat-all or treat-none scheme within the threshold between 30 and 60%. Moreover, the 2D corrected model achieved the highest benefit if the threshold probability of a patient was between 50 and 70%