Table 3.

Performance of the clinical, radiomics, and combined models in distinguishing HGPs of CRLMs.

Models		Clinical model	Radiomicstumor model	RadiomicsTLI model	Combination model
Training cohort	AUC	0.659 (0.514–0.803)	0.999 (0.997–1.000)	0.974 (0.940–1.000)	0.971 (0.927–1.000)
	ACC	0.623	0.983	0.934	0.967
	SEN	0.561	0.976	0.950	0.976
	SPE	0.750	1.000	0.927	0.950
Internal validation cohort	AUC	0.676 (0.484–0.869)	0.879 (0.741–1.000)	0.912 (0.789–1.000)	0.909 (0.785–1.000)
	ACC	0.613	0.774	0.903	0.871
	SEN	0.571	0.762	0.952	0.952
	SPE	0.700	0.800	0.800	0.700
External validation cohort	AUC	0.685 (0.563–0.807)	—	0.960 (0.919–1.000)	0.905 (0.841–0.970)
	ACC	0.567	—	0.811	0.788
	SEN	0.475	—	1.000	1.000
	SPE	0.759	—	0.414	0.345

AUC, area under curve; ACC, accuracy; SEN, sensitivity; SPE, specificity.