Table 2.

Comparison of our proposed methods (MDTL and MDTC) and other 4 state-of-the-art methods (SVM, Lasso, MTFS and M2TFS) in three binary classification problems.

AD vs. NC Classification
Method	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC	p-value
SVM	85.8	84.6	85.9	0.933	<0.0001
MDTC	88.4	87.2	88.5	0.950	<0.0001
Lasso	87.9	87.8	88.1	0.951	<0.0001
MTFS	90.7	89.5	90.8	0.966	<0.001
M2TFS	91.5	91.4	91.6	0.979	<0.005
MDTL	94.7	94.1	94.8	0.988	-
MCI vs. NC Classification
Method	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC	p-value
SVM	72.5	78.8	60.4	0.769	<0.0001
MDTC	75.8	81.4	65.1	0.810	<0.0001
Lasso	74.1	80.1	62.7	0.787	<0.0001
MTFS	78.1	83.2	68.5	0.849	<0.0005
M2TFS	78.6	83.6	69.1	0.870	<0.005
MDTL	81.5	85.8	73.3	0.882	-
pMCI vs. sMCI Classification
Method	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC	p-value
SVM	67.1	61.0	71.5	0.728	<0.0001
MDTC	69.2	63.5	73.3	0.742	<0.0001
Lasso	69.4	63.8	73.5	0.744	<0.0001
MTFS	71.0	62.7	74.9	0.757	<0.0005
M2TFS	71.4	65.8	77.8	0.768	<0.001
MDTL	73.8	69.0	77.4	0.796	-