Table 5.

Classification results of AD vs. CN on ADNI-1.

Methods	ACC	SEN	SPE	AUC	F1
Voxel+SVM	0.754	0.728	0.781	0.774	0.741
3D-CNN	0.833	0.738	0.813	0.905	0.796
Multi-slice	0.774	0.776	0.812	0.832	0.753
Multi-patch	0.808	0.710	0.793	0.890	0.767
MSA3D	0.864	0.858	0.884	0.912	0.849

All the models are trained on ADNI-2. The best results are highlighted in bold.