Table 2.

Classification Performance for our three feature extraction methods.

		Raw	SVD	HO-SVD
AD vs. NC	Accuracy	0.7104 ± 0.0816	0.7000 ± 0.1008	0.7125 ± 0.1020
	Sensitivity	0.6750 ± 0.1910	0.6750 ± 0.1985	0.7167 ± 0.1881
	Specificity	0.7222 ± 0.0845	0.7083 ± 0.1080	0.7111 ± 0.1134
	AUC	0.7611 ± 0.1162	0.7694 ± 0.1140	0.7806 ± 0.1117
AD vs. MCI	Accuracy	0.6288 ± 0.0454	0.6259 ± 0.0456	0.6894 ± 0.0612
	Sensitivity	0.5750 ± 0.1750	0.5417 ± 0.1418	0.6083 ± 0.1816
	Specificity	0.6329 ± 0.0484	0.6323 ± 0.0476	0.6956 ± 0.0691
	AUC	0.6195 ± 0.0997	0.6359 ± 0.0807	0.6520 ± 0.1086
MCI vs. NC	Accuracy	0.5311 ± 0.0529	0.5383 ± 0.0610	0.5734 ± 0.0533
	Sensitivity	0.5304 ± 0.0649	0.5384 ± 0.0707	0.5754 ± 0.0568
	Specificity	0.5375 ± 0.1677	0.5375 ± 0.1411	0.5563 ± 0.1489
	AUC	0.5585 ± 0.0892	0.5653 ± 0.0884	0.6104 ± 0.0777