Table 2.

The performance of the 2.5D CNN.

	Classifying: AD/NC Trained with: AD/NC	Classifying: MCIc/MCInc Trained with: MCIc/MCInc	Classifying: MCIc/MCInc Trained with: AD/NC	Different patch Sampling
Accuracy	88.79%	68.68%	73.04%	72.75%
Standard deviation	0.61%	1.63%	1.31%	1.20%
Confidence interval	[0.8862, 0.8897]	[0.6821, 0.6914]	[0.7265, 0.7343]	[0.7252, 0.7299]

MCIc means MCI converters. MCInc means MCI non-converters. The results were obtained with 10-fold cross validations, and averaged over 50 runs.