Preoperative prediction of WHO/ISUP grade of ccRCC using intratumoral and peritumoral habitat imaging: multicenter study

Table 2 Model performance of different machine learning algorithms in each cohort

model_name	Cohort	Accuracy	AUC	95% CI	Sensitivity	Specificity	PPV	NPV
LR	Training	0. 740	0. 852	0. 805 - 0. 898	0. 772	0. 730	0.492	0. 904
LR	Validation	0. 724	0. 774	0. 684 - 0. 863	0. 652	0. 739	0.341	0. 911
LR	Testing	0. 642	0. 720	0. 553 - 0. 888	0. 750	0. 627	0.214	0. 949
RandomForest	Training	0. 798	0. 896	0. 861 - 0. 930	0. 810	0. 794	0.571	0. 925
RandomForest	Validation	0. 806	0. 852	0. 781 - 0. 924	0. 783	0. 811	0.462	0. 947
RandomForest	Testing	0. 687	0. 835	0. 717 - 0. 953	0. 750	0. 678	0.240	0. 952
XGBoost	Training	0. 840	0. 893	0. 857 - 0. 929	0. 810	0. 850	0.646	0. 930
XGBoost	Validation	0. 776	0. 846	0. 777 - 0. 915	0. 696	0. 793	0.410	0. 926
XGBoost	Testing	0. 731	0. 841	0. 734 - 0. 948	0. 625	0. 746	0.250	0. 936
LightGBM	Training	0. 814	0. 891	0. 854 - 0. 927	0. 861	0. 798	0.591	0. 944
LightGBM	Validation	0. 828	0. 877	0. 814 - 0. 941	0. 826	0. 829	0.500	0. 958
LightGBM	Testing	0. 687	0. 847	0. 725 - 0. 969	0. 875	0. 661	0.259	0. 975

ISSN: 1470-7330