Automated Identification of Dementia Using FDG-PET Imaging

Table 2

Performance of the algorithm in binary comparisons.

	Algorithm in [18]	Algorithm in [21]	Purposed GA-MKL algorithm
AD versus normal; (86 studies)
Sensitivity	93.48%	91.30%	97.82%
Specificity	97.50%	97.50%	100%
Accuracy	91.81%	93.19%	98.89%
FTD versus normal; (83 studies)
Sensitivity	93.02%	95.35%	100%
Specificity	97.50%	95.00%	100%
Accuracy	97.64%	95.42%	100%
AD versus FTD; (89 studies)
Sensitivity	91.30%	91.11%	95.65%
Specificity	97.67%	85.71%	95.35%
Accuracy	90.83%	87.36%	94.55%

The bold font refers to the best performance obtained in each test.