Table 4.

Classification results of AD vs. CN and MCI convention on ADNI-2.

Method	AD vs. CN					pMCI vs. sMCI
	ACC	SEN	SPE	AUC	F1	ACC	SEN	SPE	AUC	F1
Voxel+SVM	0.759	0.677	0.810	0.729	0.705	0.736	0.107	0.769	0.609	0.162
3D-CNN	0.872	0.874	0.839	0.933	0.856	0.769	0.427	0.831	0.721	0.467
Multi-Slice	0.838	0.755	0.826	0.894	0.813	0.728	0.267	0.792	0.620	0.317
Multi-Patch	0.841	0.790	0.844	0.924	0.803	0.722	0.373	0.821	0.698	0.438
MSA3D	0.911	0.888	0.914	0.950	0.898	0.801	0.520	0.856	0.789	0.553

All the models are trained on ADNI-1. The best results are highlighted in bold.