Table 2.
Classification performance
| Training cohort | Test cohort | 10-fold cross-validation | |
|---|---|---|---|
| Patients vs. Controls | 96 vs. 96 | 24 vs. 44 | |
| Accuracy (%) | 72.9 | 82.3 | 80.0 |
| AUC (95% CI) | 80.6 (80.5–81.2) | 86.9 (86.6–88.8) | 83.6 (77.9–89.3) |
| Sensitivity (95% CI) | 80.2 (80.0–81.1) | 83.4 (82.6 -86.9) | 78.3 (71.4–85.3) |
| Specificity (95% CI) | 65.6 (65.3–66.3) | 81.8 (81.4–83.7) | 77.5 (70.5–84.5) |
| Positive predictive value (95% CI) | 70.0 (69.7–70.6) | 71.4 (70.4–74.1) | 77.6 (69.9–85.4) |
| Negative predictive value (95% CI) | 76.8 (67.5–77.8) | 90.0 (89.7 -92.3) | 78.4 (70.6–86.3) |
| F1-scores | 0.75 | 0.77 | 0.77 |
| MCC | 0.46 | 0.63 | 0.56 |
Performance of the multi-layer perceptron classification metric to distinguish MD from healthy controls showing the area under the curve of the Receiver Operating Curve, sensitivity, specificity, positive predictive value, negative predictive value, F1-scores and MCC. The mean F1-scores, and MCC of the 10-fold cross-validation are presented. Abbreviations: CI Confidence interval, AUC Area under the curve, MCC Matthews correlation coefficient