Table 2.

The performance of all models in predicting GAE in the training and testing cohorts.

Model	Cohort	AUC (95% CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
Clinical model	Training	0.762 (0.667–0.846)	0.748	0.721	0.780	0.800	0.696
	Testing	0.799 (0.672–0.917)	0.782	0.750	0.815	0.808	0.759
Radiomics features-combined model	Training	0.879 (0.805–0.939)	0.811	0.770	0.86	0.870	0.754
	Testing	0.724 (0.575–0.855)	0.673	0.536	0.815	0.750	0.629
Clinical–radiomics model	Training	0.886 (0.819–0.940)	0.820	0.803	0.840	0.860	0.778
	Testing	0.836 (0.707–0.937)	0.782	0.750	0.815	0.808	0.759

In the process of establishing scout models to select features, only the cross-validation performance was assessed to avoid information leakage. The bold values is optimal value.

AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value.