Table 5.
Performance and prediction based on the AD vs. CTL classification model.
| AD-like | CTL-like | Sensitivity (%) | Specificity (%) | AUC | |
|---|---|---|---|---|---|
| TRAINING SET (BASELINE, N = 348)* | |||||
| AD | 100 | 19 | 84 | 91 | 0.93 ± 0.02 |
| CTL | 10 | 100 | |||
| LONGITUDINAL DATASET (BASELINE, N = 214)** | |||||
| AD | 50 | 12 | 81 | 90 | 0.90 ± 0.03 |
| CTL | 8 | 71 | |||
| MCI-c | 11 | 2 | 85 | 65 | 0.80 ± 0.08 |
| MCI-nc | 21 | 39 | |||
| LONGITUDINAL DATASET (1 YEAR, N = 214) | |||||
| AD | 57 | 5 | 92 | 75 | 0.93 ± 0.02 |
| CTL | 20 | 59 | |||
| MCI-c | 12 | 1 | 92 | 47 | 0.75 ± 0.09 |
| MCI-nc | 32 | 28 | |||
A training set was used to train a classifier and applied on the longitudinal dataset; see Figure 1 for a description of the datasets. AD, Alzheimer disease; CTL, healthy control. MCI, mild cognitive impairment; MCI-c, MCI converters; MCI-nc, MCI non-converters; AUC, area under the receiver operating characteristic curve.
*
Training model using cross-validation.
**
Values recalculated from the baseline cross-validated model.