Table 2.

Diagnostic performance of models constructed by Resnet-50 on the training and validation cohorts

Group	Models	AUC (95%CI)	Accuracy	Sensitivity	Specificity	PPV	NPV
Training set	T1WI	0.804 (0.756–0.851)	0.764	0.419	0.928	0.735	0.771
	T2WI	0.817 (0.770–0.863)	0.760	0.419	0.923	0.720	0.770
	CE-T1WI	0.845 (0.802–0.889)	0.790	0.558	0.901	0.727	0.811
	DLS	0.981 (0.965–0.997)	0.929	0.826	0.978	0.947	0.922
Internal validation set	T1WI	0.705 (0.621–0.755)	0.626	0.865	0.513	0.457	0.889
	T2WI	0.741 (0.661–0.821)	0.565	0.865	0.423	0.416	0.868
	CE-T1WI	0.769 (0.692–0.846)	0.696	0.757	0.667	0.519	0.853
	combined	0.860 (0.797–0.924)	0.774	0.811	0.756	0.612	0.894
External validation set	T1WI	0.689 (0.611–0.768)	0.619	0.800	0.449	0.578	0.705
	T2WI	0.747 (0.674–0.821)	0.702	0.846	0.565	0.647	0.796
	CE-T1WI	0.734 (0.659–0.809)	0.642	0.600	0.681	0.639	0.644
	combined	0.803 (0.735–0.870)	0.754	0.708	0.797	0.767	0.743

AUC, area under the curve; CI, confidence interval; DL, deep learning; NPV, negative predictive value; PPV, positive predictive value