Table 1.
Classification performance.
| Classifier | AUC | Accuracy (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | F1 Score |
|---|---|---|---|---|---|---|---|
| CNN: Status Balanced | 0.95 | 89 | 86 | 92 | 91 | 86 | 88 |
| CNN: Location Balanced | 0.86 | 80 | 72 | 89 | 87 | 76 | 78 |
| CNN: xMask | 0.86 | 76 | 76 | 76 | 76 | 76 | 76 |
| RF: Status Balanced | 0.90 | 90 | 84 | 95 | 94 | 86 | 89 |
| RF: Location Balanced | 0.65 | 65 | 70 | 60 | 63 | 67 | 67 |
| Expert 1 | 0.86 | 86 | 80 | 93 | 92 | 82 | 85 |
| Expert 2 | 0.75 | 74 | 50 | 98 | 97 | 66 | 66 |
Classification results are displayed in percentages. The optimal threshold for the three CNNs was selected by maximizing Youden’s Index. RF: Random Forest.