Table 3.
The training and testing size, sensitivity, specificity, and the F1-Score for each of the models, i.e. VGG16, DenseNet161, and ResNet18
| VGG16 | Normal | Tuberculosis | Pneumonia affected |
|---|---|---|---|
| Total Images | 492 | 382 | 638 |
| Training size | 388 | 303 | 500 |
| Testing size | 104 | 79 | 138 |
| Sensitivity | 100% | 98% | 92% |
| Specificity | 97% | 97% | 99% |
| F1 Score | 0.97 | 0.96 | 0.95 |
| Densenet161 | Pneumonia | COVID-19 | |
| Total Images | 750 | 1175 | |
| Training size | 600 | 1025 | |
| Testing size | 150 | 235 | |
| Sensitivity | 99% | 98% | |
| Specificity | 98% | 99% | |
| F1 Score | 0.98 | 0.99 | |
| ResNet-18 | Medium | Mild | Severe |
| Total Images | 30 | 26 | 26 |
| Training size | 21 | 18 | 18 |
| Testing size | 9 | 8 | 8 |
| Sensitivity | 77% | 100% | 50% |
| Specificity | 75% | 94% | 94% |
| F1 Score | 0.7 | 0.94 | 0.61 |