Table 4.
Class-wise performance metrics as achieved on different models used in this implementation
| Model | Class | Precision | Recall | F1-Score | Support | Accuracy (%) | Sensitivity (%) | Specificity (%) |
|---|---|---|---|---|---|---|---|---|
| Covid | 0.6601 | 0.994 | 0.7933 | 168 | ||||
| StackNet-DenVIS | Non-Covid | 0.9993 | 0.9461 | 0.972 | 1596 | 95.07 | 99.4 | 94.61 |
| Macro-average | 0.8297 | 0.9701 | 0.8827 | 1764 | ||||
| Weighted-average | 0.967 | 0.9507 | 0.955 | 1764 | ||||
| Covid | 0.7703 | 0.9583 | 0.9825 | 168 | ||||
| VGG19 bn | Non-Covid | 0.9955 | 0.9699 | 0.8541 | 1596 | 96.88 | 95.83 | 96.99 |
| Macro-average | 0.8829 | 0.9641 | 0.9183 | 1764 | ||||
| Weighted-average | 0.9741 | 0.9688 | 0.9703 | 1764 | ||||
| Covid | 0.6653 | 0.9583 | 0.7854 | 168 | ||||
| SE-ResNeXt50-32 × 4d | Non-Covid | 0.9954 | 0.9492 | 0.9718 | 1596 | 95.01 | 95.83 | 94.92 |
| Macro-average | 0.8303 | 0.9538 | 0.8786 | 1764 | ||||
| Weighted-average | 0.964 | 0.9501 | 0.954 | 1764 | ||||
| Covid | 0.5189 | 0.9821 | 0.679 | 168 | ||||
| Inception ResNet v2 | Non-Covid | 0.9979 | 0.9041 | 0.9487 | 1596 | 91.16 | 98.21 | 90.41 |
| Macro-average | 0.7584 | 0.9431 | 0.8139 | 1764 | ||||
| Weighted-average | 0.9523 | 0.9116 | 0.923 | 1764 | ||||
| Covid | 0.3756 | 0.881 | 0.5267 | 168 | ||||
| DenseNet-121 | Non-Covid | 0.9854 | 0.8459 | 0.9103 | 1596 | 84.92 | 88.1 | 84.59 |
| Macro-average | 0.6805 | 0.8634 | 0.7185 | 1764 | ||||
| Weighted-average | 0.9273 | 0.8492 | 0.8738 | 1764 |