Table 3.

Diagnostic performance of the clinical factor model, the radiomics signature, and the radiomics nomogram.

Model	Group	AUC  (95% CI)	Sensitivity	Specificity	Accuracy	PPV	NPV	F1-score
Clinical factor model	Training cohort	0.803 (0.723–0.876)	0.891 (98/110)	0.633 (19/30)	0.836 (117/140)	0.899 (98/109)	0.613 (19/31)	0.895
	Internal validation cohort	0.763 (0.603–0.841)	0.875 (42/48)	0.571 (8/14)	0.806 (50/62)	0.875 (42/48)	0.571 (8/14)	0.875
	External validation cohort	0.817 (0.690–0.833)	0.533 (16/30)	1 (13/13)	0.674 (29/43)	1 (16/16)	0.481 (13/27)	0.696
Radiomics signature	Training cohort	0.813 (0.743–0.877)	0.736 (81/110)	0.7 (21/30)	0.729 (102/140)	0.9 (81/90)	0.42 (21/50)	0.81
	Internal validation cohort	0.787 (0.606–0.825)	0.667 (32/48)	0.786 (11/14)	0.694 (43/62)	0.914 (32/35)	0.407 (11/27)	0.771
	External validation cohort	0.718 (0.531–0.778)	0.633 (19/30)	0.692 (9/13)	0.651 (28/43)	0.826 (19/23)	0.45 (9/20)	0.717
Radiomics nomogram	Training cohort	0.887 (0.827–0.940)	0.818 (90/110)	0.8 (24/30)	0.814 (114/140)	0.938 (90/96)	0.545 (24/44)	0.874
	Internal validation cohort	0.885 (0.816–0.922)	0.75 (36/48)	1 (14/14)	0.806 (50/62)	1 (36/36)	0.538 (14/26)	0.857
	External validation cohort	0.851 (0.720–0.919)	0.8 (24/30)	0.846 (11/13)	0.814 (35/43)	0.923 (24/26)	0.647 (11/17)	0.857

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