| AUC | Accuracy | Sensitivity (recall) | Specificity | PPV (precision) | NPV | f1-score |
---|---|---|---|---|---|---|---|
Inception-ResNet-V2 | 91.8% (87.3–96.3%) | 95.18% | 90.35% | 100% | 100% | 83.58% | 96.6% |
VGG-16 | 81.3% (75.2–87.4%) | 80.86% | 86% | 75.71% | 83.5% | 79.1% | 84.7% |
InceptionV3 | 89.4% (84.4–94.3%) | 90.86% | 89.91% | 91.8% | 95.15% | 83.58% | 92.46% |
Model ensemble | 89.4% (84.4–94.3%) | 90.86% | 89.91% | 91.8% | 95.15% | 83.58% | 92.46% |