Table 4.
Random Forest performance validation for predicting whether MCI subjects will progress to AD or not (tenfold cross-validation; Second Layer).
| Modalities used | Precision (%) | Recall (%) | Accuracy (%) | F1-score (%) | AUC |
|---|---|---|---|---|---|
| All | 87.12 ± 1.52 | 81.31 ± 2.00 | 84.18 ± 1.77 | 83.21 ± 2.43 | 84.17 ± 1.99 |
| CFA | 82.14 ± 1.40 | 84.19 ± 1.90 | 82.16 ± 1.60 | 83.15 ± 1.50 | 82.15 ± 1.50 |
| MRI | 75.23 ± 1.88 | 72.25 ± 1.59 | 71.18 ± 2.01 | 72.17 ± 1.92 | 71.18 ± 1.89 |
| PET | 68.22 ± 2.22 | 68.53 ± 1.98 | 68.25 ± 1.99 | 66.39 ± 1.99 | 68.24 ± 2.01 |
| Genetics | 73.11 ± 1.73 | 68.36 ± 1.72 | 70.14 ± 1.79 | 69.24 ± 1.89 | 70.13 ± 1.80 |
| MH | 58.16 ± 4.60 | 52.22 ± 4.90 | 55.15 ± 3.73 | 54.19 ± 3.76 | 55.15 ± 3.76 |
| CFA + MRI | 83.11 ± 2.31 | 84.22 ± 2.20 | 83.14 ± 2.51 | 83.15 ± 2.26 | 83.14 ± 2.22 |
| CFA + PET | 85.17 ± 1.70 | 84.29 ± 2.90 | 84.19 ± 1.90 | 84.21 ± 2.10 | 84.19 ± 1.90 |
| CFA + Genetics | 82.14 ± 1.91 | 81.27 ± 1.98 | 81.16 ± 1.93 | 81.17 ± 1.96 | 81.16 ± 1.95 |
| CFA + MH | 84.16 ± 3.77 | 82.23 ± 4.60 | 82.17 ± 3.80 | 83.18 ± 3.33 | 82.17 ± 3.80 |
| CFA + PET + MRI | 86.09 ± 2.10 | 84.23 ± 2.30 | 84.11 ± 2.15 | 85.14 ± 2.22 | 85.11 ± 2.15 |
| CFA + PET + Genetics | 90.11 ± 1.50 | 83.21 ± 2.21 | 86.08 ± 1.04 | 85.11 ± 2.00 | 86.08 ± 1.05 |
| CFA + PET + MH | 86.09 ± 3.45 | 84.17 ± 4.20 | 84.08 ± 3.71 | 85.09 ± 4.01 | 84.08 ± 3.72 |
| CFA + PET + Genetics + MRI* | 88.07 ± 0.70 | 86.13 ± 1.30 | 87.08 ± 0.80 | 87.09 ± 0.90 | 87.08 ± 0.80 |
| CFA + PET + Genetics + MH | 86.09 ± 3.51 | 86.13 ± 4.70 | 86.08 ± 3.32 | 86.08 ± 3.99 | 86.08 ± 3.36 |
BA: Balanced accuracy. Asterisk ( ∗): is the subset of features with the best predictive performance. Performance: Mean ± standard deviation.