Table 2.
Accuracy of the previous methods and the proposed method for different AD patient groups
| Method | Classification Model Accuracy (%) | |||
|---|---|---|---|---|
| AD/NC | AD/MCI | NC/MCI | MCIc/MCInc | |
| Chupin et al., 2009 [8] | 80.51 | 73.48 | 71.94 | 64.21 |
| Ahmed et al., 2015 [11] | 86.40 | 74.51 | 76.29 | 68.72 |
| Khedher et al., 2015 [12] | 88.96 | 84.59 | 82.41 | 70.11 |
| Dai et al., 2013 [9] | 90.81 | 85.92 | 81.92 | 71.04 |
| Suk et al., 2014 [10] | 93.05 | 88.98 | 83.67 | 72.86 |
| Liu et al., 2018 [13] | 95.24 | 90.85 | 86.35 | 74.28 |
| Proposed method (the best result among 20 runs) | 96.50 | 91.25 | 87.65 | 78.49 |
| Proposed method (at 90% confidence level of t-test) | 95.80 | 90.63 | 86.47 | 77.42 |