Table 2.

Comparison of the predictive performance of clinical model, radiomics models, and joint models

	AUC	95% CI low	95% CI up	Best threshold	Specificity	Sensitivity	Accuracy
Training set
Clinical model	0.738	0.688	0.788	0.643	0.683	0.708	0.698
CT_RF	0.688	0.637	0.739	0.635	0.714	0.547	0.614
CT joint model	0.773	0.727	0.818	0.638	0.795	0.638	0.701
PET_RF	0.666	0.613	0.719	0.525	0.615	0.670	0.649
PET joint model	0.743	0.694	0.792	0.533	0.590	0.794	0.713
PET/CT_RF	0.698	0.647	0.749	0.549	0.739	0.564	0.634
PET/CT joint model	0.760	0.713	0.807	0.573	0.683	0.733	0.713
Testing set
Clinical model	0.681	0.577	0.782	0.538	0.634	0.686	0.667
CT_RF	0.726	0.629	0.822	0.596	0.756	0.643	0.685
CT joint model	0.723	0.628	0.818	0.504	0.683	0.671	0.676
PET_RF	0.678	0.572	0.785	0.504	0.659	0.643	0.649
PET joint model	0.703	0.601	0.806	0.575	0.634	0.686	0.667
PET/CT_RF	0.704	0.603	0.804	0.426	0.561	0.800	0.712
PET/CT joint model	0.730	0.633	0.828	0.491	0.585	0.786	0.712

AUC area under the curve; CI confidence interval; RF random forest