Table 3.

Classification results per group using the contrast-enhanced T1-weighted images. AUC: area under the receiver operator characteristic curve. Sens.: sensitivity. Spec.: specificity. Acc.: accuracy.

Number of	Training data				Independent test data
features	AUC	Sens.	Spec.	Acc.	AUC	Sens.	Spec.	Acc.
1	0.7943	0.6853	0.7014	0.6935	0.7308	0.7533	0.4417	0.5975
2	0.8247	0.6853	0.8290	0.7586	0.7525	0.6842	0.7475	0.7158
3	0.8701	0.7037	0.8153	0.7606	0.7314	0.6867	0.7067	0.6967
4	0.8805	0.7241	0.8294	0.7778	0.7329	0.7108	0.7075	0.7092
5	0.8979	0.7590	0.8529	0.8069	0.8035	0.8142	0.6325	0.7233
6	0.9225	0.7788	0.8884	0.8347	0.8192	0.8167	0.7225	0.7696
7	0.9388	0.8002	0.9041	0.8532	0.8128	0.8017	0.7175	0.7596
8	0.9345	0.7978	0.8980	0.8489	0.8142	0.7992	0.7083	0.7538
9	0.9435	0.8273	0.9059	0.8674	0.8117	0.8000	0.7383	0.7692
10	0.9347	0.8114	0.8965	0.8548	0.8028	0.7750	0.7650	0.7700
11	0.9308	0.8129	0.8912	0.8528	0.8114	0.7725	0.7825	0.7775
12	0.9322	0.8163	0.8902	0.8540	0.7930	0.7283	0.7808	0.7546
13	0.9272	0.8137	0.8857	0.8504	0.7837	0.7408	0.7733	0.7571
14	0.9299	0.8198	0.8800	0.8505	0.7632	0.7175	0.7475	0.7325
15	0.9318	0.8282	0.8851	0.8572	0.7707	0.7333	0.7442	0.7388