Table 3.
Classification performance comparison between the proposed method and related methods for distinguishing HC, MCI and AD
| Methods | Accuracy (mean + SD) |
|---|---|
| SVM (SNP) | 0.561 ± 0.026 |
| SVM (FreeSurfer) | 0.573 ± 0.012 |
| SVM (VBM) | 0.541 ± 0.032 |
| SVM (PET) | 0.535 ± 0.026 |
| SVM (all) | 0.575 ± 0.019 |
| HML (all) | 0.638 ± 0.019 |
| SVM ℓ∞ MKL method | 0.624 ± 0.031 |
| SVM ℓ1 MKL method | 0.593 ± 0.042 |
| SVM ℓ2 MKL method | 0.561 ± 0.037 |
| LSSVM ℓ∞ MKL method | 0.614 ± 0.031 |
| LSSVM ℓ1 MKL method | 0.585 ± 0.018 |
| LSSVM ℓ2 MKL method | 0.577 ± 0.033 |
| Our method (SNP) | 0.673 ± 0.021 |
| Our method (FreeSurfer) | 0.689 ± 0.029 |
| Our method (VBM) | 0.669 ± 0.031 |
| Our method (PET) | 0.621 ± 0.028 |
| Our method | 0.726 ± 0.032 |