Table 8.
We compare our proposed COVID-SRWCNN model with state-of-the-art COVID-19 image-based diagnosis models.
| Literature | Architecture | Performance (%) |
|---|---|---|
| Wang et al. [56] | 2D CNN | 82.9 (ACC) |
| Shi et al. [44] | Random forest-based CNN | 87.9 (ACC) 83.3 (SEN) 90.7 (SPE) |
| Chen et al. [38] | 2D Unet ++ | 95.2 (ACC) 100.0 (SEN) 93.6 (SPE) |
| Li et al. [42] | 2D ResNet 50 | 90.0 (SEN) 96.0 (SPE) |
| Song et al. [49] | 2D ResNet 50 | 86.0 (ACC) |
| Jin et al. [45] | 2D Unet++ and 2D CNN | 97.4 (SEN) 92.2 (SPE) |
| Xu et al. [47] | 2D CNN | 86.7 (ACC) |
| Jin et al. [46] | 2D CNN | 94.1 (SEN) 95.5 (SPE) |
| Wang et al. [48] | 3D ResNet and attention | 93.3 (ACC) 87.6 (SEN) 95.5 (SPE) |
| Zhang et al. [37] | 2D Unet and 2D CNN | 90.7 (SEN) 90.7 (SPE) |
| COVID-SRWCNN | Super-Resolution CNN and Wavelet | 99.79 (ACC) 99.78 (SEN) 99.86 (SPE) 98.96 (AUC) 98.98 (PRE) |