Table 3.

Performance of clinico-radiological, CE-T1, T2, and CE-T1+T2 models, and nomogram

Model	Performance	AUC (95% CI)	ACC	SEN	SPE	p value	Cut-off
Clinico-radiological	Training set	0.846 (0.831–0.861)	0.763	0.765	0.761	1.25E-9	0.472
	Test set	0.828 (0.812–0.844)	0.773	0.823	0.686	1.63E-8	0.472
CE-T1	Training set	0.852 (0.837–0.868)	0.753	0.851	0.660	2.33E-9	0.266
	Test set	0.826 (0.804–0.844)	0.804	0.800	0.807	1.07E-7	0.266
T2	Training set	0.768 (0.748–0.787)	0.711	0.809	0.620	5.71E-6	-0.091
	Test set	0.733 (0.712–0.754)	0.680	0.629	0.710	1.46E-4	-0.091
CE-T1+T2	Training set	0.869 (0.855–0.884)	0.753	0.851	0.660	3.80E-10	0.134
	Test set	0.803 (0.784–0.821)	0.791	0.771	0.790	8.16E-7	0.134
Nomogram	Training set Test set	0.899 (0.887–0.911) 0.871 (0.857–0.885)	0.814 0.794	0.936 0.857	0.700 0.758	1.31E-11 1.51E-9	-0.732 -0.732

The best performance in the test cohort is indicated in bold font. The cutoff values were calculated using the xtile function in R

AUC area under the curve, ACC accuracy, SEN sensitivity, SPE specificity, CE-T1 contrast-enhanced T1 weighted MR image, T2 T2 weighted MR image