Table 6.
Comparison of classification performance of different methods.
| Article | Method | MCInc/MCIc | Scans | ACC (%) | SEN (%) | SPE (%) | AUC |
|---|---|---|---|---|---|---|---|
| Cui et al., 2011 | Multivariate predictors (MRI, CSF, and NM scores) | 87/56 | baseline | 67.1 | 96.4 | 48.3 | 0.796 |
| Ye et al., 2012 | SLR+SS (MRI, genetic, and cognitive measures) | 177/142 | baseline | – | – | – | 0.859 |
| Eskildsen et al., 2013 | Patterns of cortical thinning | 134/122 | 6 months | 75.8 | 75.4 | 76.1 | 0.809 |
| 134/123 | 12 months | 72.9 | 75.8 | 70.2 | 0.762 | ||
| Raamana et al., 2015 | Thickness network fusion | 130/56 | baseline | 64.0 | 65.0 | 64.0 | 0.680 |
| Proposed | Combination of MRI and thickness network | 83/76 | baseline | 66.0 | 55.3 | 75.9 | 0.735 |
| 83/61 | 6 months | 76.4 | 65.6 | 84.3 | 0.813 | ||
| 83/63 | 12 months | 74.7 | 65.1 | 81.9 | 0.785 | ||
| 83/42 | 18 months | 73.9 | 70.5 | 77.1 | 0.773 |
The best multivariate predictors of MCI conversion are shown for each study.
ACC, accuracy; SEN, sensitivity; SPE, specificity; AUC, area under the curve; CSF, Cerebrospinal Fluid; NMs, neuropsychological and functional measures; SLR+SS, sparse logistic regression with stability selection.