Table 7.
Method | Subjects | Modality | CN vs. AD |
CN vs. pMCI |
||||||
---|---|---|---|---|---|---|---|---|---|---|
ACC | SEN | SPE | AUC | ACC | SEN | SPE | AUC | |||
Liu et al., 2018 | 93 AD + 204 MCI + 100 CN | MRI | 84.97% | 82.65% | 87.37% | 90.63% | 77.84% | 76.81% | 78.59% | 82.72% |
PET | 88.08% | 90.70% | 85.98% | 94.51% | 78.41% | 77.94% | 78.70% | 85.96% | ||
Both | 93.26% | 92.55% | 93.94% | 95.68% | 82.95% | 81.08% | 84.31% | 88.43% | ||
Proposed method | 465 AD + 567 MCI + 480 CN | MRI | 81.19% | 79.27% | 83.45% | 83.67% | – | – | – | – |
PET | 89.11% | 90.24% | 87.77% | 92.69% | – | – | – | – | ||
Both | 90.10%1 | 90.85% | 89.21% | 90.84% | 82.38%2 | 87.20% | 72.50% | 81.64% | ||
Both | – | – | – | – | 87.46%3 | 90.73% | 80.61% | 87.61% |
1Using B1 protocol, the CN vs. AD training set and the CN vs. AD testing set. 2Using B2 protocol, the CN vs. pMCI training set and the CN vs. pMCI testing set. 3Using B1 protocol, the CN vs. AD training but the CN vs. pMCI testing set. See Table 9 for reference. The best results were indicated in bold.