Table 3.

Predictive performances of radiomics nomogram, Rad-score, conventional CT model, and judgment by radiologists in the training and validation cohorts.

Model	Training cohort				Validation cohort
	AUC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Accuracy	AUC (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Accuracy
Radiomics nomogram	0.980 (0.963–0.993)	0.870 (0.779–0.928)	0.983 (0.897–0.999)	0.914	0.992 (0.969–1.000)	0.958 (0.769–0.998)	0.933 (0.660–0.997)	0.949
Rad-score	0.909 (0.864–0.948)	0.946 (0.872–0.980)	0.746 (0.613–0.846)	0.868	0.953 (0.893–0.997)	0.917 (0.715–0.985)	0.867 (0.584–0.977)	0.897
Conventional CT	0.917 (0.882–0.949)	0.783 (0.682–0.859)	0.898 (0.785–0.958)	0.828	0.868 (0.759–0.944)	0.583 (0.369–0.772)	1.000 (0.747–1.000)	0.744
Radiologist	NA	0.859 (0.767–0.920)	0.373 (0.253–0.509)	0.669	NA	0.792 (0.573–0.921)	0.400 (0.175–0.671)	0.641