Table 1.

Evaluation on test data utilizing two hold-out strategies when splitting the training and test sets.

	Data Split by Subject (Brain Hold-Out)			Data Split by Slice (Slice Hold-Out)
Binary Classes	Accuracy (%)	Sensitivity (%)	Specificity (%)	Accuracy (%)	Sensitivity (%)	Specificity (%)
EMCI vs. LMCI	53.08	23.96	42.78	83.90	79.39	88.47
AD vs. CN	48.53	23.20	33.00	73.18	48.21	98.16
CN vs. EMCI	40.04	25.14	20.07	86.90	85.75	88.00
CN vs. LMCI	56.97	27.17	40.54	84.12	75.18	93.07
EMCI vs. AD	45.24	28.00	24.26	79.55	82.36	76.74
LMCI vs. AD	44.12	2.62	52.56	84.58	95.69	73.48
MCI vs. EMCI	46.22	27.51	23.76	89.86	86.43	93.33