Table 2.

Diagnostic performance of the HCR, CM, TLRS, and TLRN in the training and validation sets.

Model		AUC(95% CI)	Sensitive	Specificity	Accuracy	PPV	NPV
Training cohort	CM	0.894 (0.827–0.960)	0.850	0.800	0.836	0.919	0.667
	TLRS	0.986 (0.942–0.999)	0.975	0.967	0.973	0.987	0.936
	TLRN	0.989 (0.947–1.000)	0.975	0.967	0.973	0.987	0.936
Internal validation cohort	CM	0.820 (0.718–0.921)	0.720	0.796	0.770	0.936	0.667
	TLRS	0.904 (0.814–0.961)	0.980	0.720	0.892	0.872	0.947
	TLRN	0.958 (0.883–0.991)	0.857	0.960	0.891	0.976	0.774
External validation cohort 1	CM	0.816 (0.736–0.880)	0.827	0.687	0.772	0.805	0.717
	TLRS	0.834 (0.756–0.895)	0.986	0.729	0.886	0.850	0.972
	TLRN	0.867 (0.794–0.922)	0.987	0.729	0.886	0.851	0.972
External validation cohort 2	CM	0.866 (0.795–0.919)	0.926	0.640	0.817	0.807	0.842
	TLRS	0.894 (0.828–0.941)	0.852	0.900	0.872	0.932	0.790
	TLRN	0.921 (0.860–0.961)	0.926	0.820	0.886	0.893	0.872

AUC, area under curve; CI, confidence interval; CM, clinical model; TLRS, transfer learning radiomics signature; TLRN, transfer learning radiomics nomogram.