Table 4.

Detailed diagnosis performance of models in all datasets.

	Model	Accuracy (95% CI)	Sensitivity	Specificity	PPV	NPV
Training set	Radiomics	0.86 (0.80-0.91)	0.87	0.86	0.86	0.87
	Conventional	0.89 (0.84-0.94)	0.83	0.95	0.95	0.85
	Nomogram	0.90 (0.85-0.94)	0.93	0.88	0.89	0.93
Test set	Radiomics	0.82 (0.71-0.90)	0.76	0.87	0.84	0.80
	Conventional	0.83 (0.73-0.91)	0.74	0.92	0.89	0.80
	Nomogram	0.88 (0.78-0.94)	0.85	0.89	0.88	0.87
External validation set	Radiomics	0.83 (0.70-0.92)	0.83	0.83	0.79	0.86
	Conventional	0.79 (0.65-0.89)	0.70	0.86	0.80	0.78
	Nomogram	0.77 (0.63-0.87)	0.87	0.69	0.69	0.87

PPV, positive predict value; NPV, negative predict value.

The cutoff of radiomics model is -0.1155177, the cutoff of conventional model is 0.6482431, the cutoff of nomogram model is -0.6291612.