Table 10.
Comparison with state-of-the-art approaches.
| Author | Architecture | Accuracy | F1 | Recall | Precision |
|---|---|---|---|---|---|
| Matsuyama, E. [17] | ResNet50 + wavelet coefficients | 92.2% | 91.5% | 90.4% | / |
| Loey, M. [18] | ResNet50 + augumentation + CGAN | 82.91% | / | 77.66% | / |
| Do, C. [19] | Modified DenseNet201 | 85% | / | 79% | 91% |
| Polsinelli, M. [20] | Modified SqueezeNet | 85.03% | 86.20% | 87.55% | 85.01% |
| Panwar, H. [45] | Modified VGG19 | 94.04 | |||
| Mishra, A. [46] | Modified DenseNet121, ResNet50, VGG16, InceptionV3 and DenseNet201 | 88.3% | 86.7% | 90.15% | |
| Ko, H. [21] | Modified VGG16, ResNet-50, Inception-v3, and Xception | 96.97% | |||
| Maghdid, H. [22] | Modified Alexnet, A self-build CNN | 94.1% | 100% | ||
| Arora, V. [47] | Modified MobileNet, DenseNet121, ResNet50, VGG16, InceptionV3 and XceptionNet |
94.12% | 96.11% | 96.11% | 96.11% |
| Alshazly. H. [48] | CovidResNet and CovidDenseNet | 93.87% | 95.70 | 92.49 | 99.13% |
| Yu, Z. [49] | Modified InceptionV3, ResNet50, ResNet-101, DenseNet201 |
95.34% | |||
| Jaiswal, A. [50] | Modified DenseNet201 | 96.25% | 96.29% | 96.29% | 96.29% |
| Sanagavarapu, S. [51] | Ensembled ResNets | 87% | 84% | 81% | 91% |
| Song, J. [52] | A large-scale bi-directional generative adversarial network | 92% | |||
| Sarker, L [53] | Modified Densenet121 | 96.49% | 96% | 96% | 96% |
| Shan, F. [54] | VB-Net | 91.6% | |||
| Wang, S. [55] | Modified DenseNet | 85% | 90% | 79% | |
| Gozes, O. [56] | Modified ResNet50 | 94% | |||
| Wang, S. [57] | Modified Inception | 79.3% | 63% | 83% | |
| Li, L. [58] | Modified RestNet50 | 90% | |||
| Proposed | EDNC | 97.75% | 97.75% | 97.95% | 97.55% |