Table 4. Classification performance of SVM-RFE.

Features	sMCI-NC							cMCI-NC						cMCI-sMCI
	Original dimension	Optimized dimension	AUC	Acc (%)	Sen (%)	Spe (%)		Optimized dimension	AUC	Acc (%)	Sen (%)	Spe (%)		Optimized dimension	AUC	Acc (%)	Sen (%)	Spe (%)
Cortical thickness	68	58	0.6989	66.98	59.22	74.31		20	0.8753	81.19	80.37	81.56		63	0.7582	73.4	73.87	72.82
Gray volume	68	62	0.639	61.79	83.49	38.83		44	0.8123	74.31	77.06	71.56		63	0.7326	71.2	70.18	68.69
Mean curvature	68	45	0.6479	60.38	59.22	61.47		48	0.7344	68.18	68.47	67.89		43	0.6645	65.4	70.06	55.62
Surface area	68	43	0.6029	61.32	56.31	66.06		61	0.6197	59.09	63.06	55.05		35	0.6437	64.5	71.17	57.28
hippocampal sub-regions	16	4	0.681	64.62	61.17	67.89		5	0.8511	78.96	78.13	72.63		8	0.726	70.3	70.86	70.63
Combined cortical features	272	223	0.7374	70.75	67.96	73.39		205	0.8863	83.41	85.13	79.36		203	0.7869	75.2	75.36	76.55
Cortical features combined with volume of hippocampal subregions	288	235	0.754	71.86	70.12	73.62		235	0.905	84.64	85.32	81.33		233	0.797	76.9	75.66	78.14

AUC, area under the curve; Acc, accuracy; Sen, sensitivity; Spe, specificity; sMCI, stable mild cognitive impairment; cMCI, converted mild cognitive impairment; NC, normal control; SVM-RFE, support vector machine-recursive feature elimination.