Table 2.
Comparative performance metrics of the different deep learning networks performing classification of pneumonia, TB, COVID-19, and healthy cases. Boldface indicates the best metric among the networks.
| DXR1 dataset: pneumonia and healthy | ||||
|---|---|---|---|---|
| Metric | DenResCov-19 | DenseNet-121 | ResNet-50 | Inception-V3 |
| Recall (%) | 98.12 | 97.80 | 97.71 | 93.32 |
| Precision (%) | 98.31 | 94.62 | 95.01 | 90.10 |
| AUC-ROC (%) | 99.60 | 99.10 | 98.95 | 92.80 |
| F1 (%) | 98.21 | 96.27 | 96.34 | 91.68 |
| DXR2 dataset: COVID-19, pneumonia and healthy | ||||
|---|---|---|---|---|
| Metric | DenResCov-19 | DenseNet-121 | ResNet-50 | VGG-16 |
| Recall (%) | 89.38 | 83.54 | 83.53 | 99.83 |
| Precision (%) | 85.28 | 77.45 | 73.35 | 33.38 |
| AUC-ROC (%) | 96.51 | 93.2 | 92.39 | 50.07 |
| F1 (%) | 87.29 | 80.37 | 78.11 | 49.51 |
| DXR3 dataset: COVID-19, pneumonia, tuberculosis and healthy | ||||
|---|---|---|---|---|
| Metric | DenResCov-19 | DenseNet-121 | ResNet-50 | VGG-16 |
| Recall (%) | 59.28 | 57.71 | 56.66 | 66.53 |
| Precision (%) | 79.56 | 74.87 | 74.00 | 26.53 |
| AUC-ROC (%) | 91.77 | 89.49 | 92.12 | 53.11 |
| F1 (%) | 68.09 | 65.17 | 64.17 | 38.00 |
| DXR4 dataset: COVID-19, pneumonia, tuberculosis and healthy | ||||
|---|---|---|---|---|
| Metric | DenResCov-19 | DenseNet-121 | ResNet-50 | VGG-16 |
| Recall (%) | 69.7 | 62.70 | 62.00 | 93.69 |
| Precision (%) | 82.90 | 79.35 | 78.60 | 27.17 |
| AUC-ROC (%) | 95.00 | 91.00 | 93.21 | 54.99 |
| F1 (%) | 75.75 | 70.07 | 69.51 | 42.13 |