Table 6.
Comparison of the state-of-art models.
| Methods | Dataset | # of classes | Acc (%) | Se (%) | Sp (%) |
|---|---|---|---|---|---|
| DarkCovidNet [9] | Public | 3 | 87.02 | 92.18 | 89.96 |
| COVIDiagnosis-Net [8] | Public | 3 | 98.26 | 99.13 | – |
| The pretrained CNNs [43] | Public | 3 | 93.48 | 92.85 | 98.75 |
| COVID-Net [44] | Public | 3 | 92.64 | 91.37 | 95.76 |
| Deep features, ResNet-50, SVM [45] | Public | 2 | 95.38 | – | – |
| Deep CNNs [46] | Public | 2 | 90.00 | 100 | 80.00 |
| Deep CNN, ResNet-50 [47] | Public | 2 | 98.00 | – | – |
| DRE-Net, deep CNN [48] | Private dataset | 2 | 86.00 | 96.00 | – |
| Deep CNN, Inception, transfer learning [49] | Private dataset | 2 | 89.50 | 87.00 | 88.00 |
| nCOVnet, transfer learning, deep CNN [31] | Public | 2 | 88.10 | 97.62 | 89.13 |
| A novel CNN model, training from scratch strategy, deep feature extraction, SVM | Public | 3 | 98.97 | 89.39 | 99.75 |