Table 3.

Performance of different classification approaches in multitasking classification.

Model	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC
AD and HC groups
GWAS analysis	71.38 ± 0.63	63.13 ± 2.87	85.59 ± 6.66	0.744
CNN model	92.45 ± 8.13	93.87 ± 12.26	90.00 ± 15.97	0.915
ResNet18	97.96 ± 1.71	97.42 ± 3.16	98.89 ± 1.36	0.980
ResNet34	98.78 ± 1.50	98.39 ± 2.50	99.44 ± 1.11	0.981
MCI and HC groups
GWAS analysis	56.99 ± 1.55	96.08 ± 13.92	5.94 ± 21.65	0.510
CNN model	87.47 ± 16.64	99.57 ± 0.85	71.67 ± 38.75	0.852
ResNet18	97.59 ± 3.73	100.00 ± 0.00	94.44 ± 8.61	0.966
ResNet34	99.52 ± 0.60	99.57 ± 0.85	99.44 ± 1.11	0.986
AD and MCI groups
GWAS analysis	58.97 ± 0.00	72.18 ± 0.01	41.54 ± 0.01	0.569
CNN model	86.42 ± 16.02	97.42 ± 4.40	71.91 ± 39.21	0.840
ResNet18	97.80 ± 1.24	97.74 ± 2.41	97.87 ± 3.30	0.972
ResNet34	98.90 ± 1.78	100.00 ± 0.00	97.45 ± 4.13	0.981

The methods are conducted with crossvalidation, and their results are given as mean ± standard deviation.