Table 1.
Overview of studies using deep learning approaches for COVID-19 detection.
| Ref | CNN Model | Data Sources | Accuracy (%) | Limitations |
|---|---|---|---|---|
| [15] | nCOVnet | Cohen et al. [25] | 97.97 | High execution time |
| [16] | ResNet50 | Radiography Database [26] | 99.17 | Imbalanced data |
| [17] | ResNet18 | CT scan images [27] | 99.60 | Overfitting issue |
| ResNet50 | 99.20 | |||
| ResNet101 | 99.30 | |||
| SqueezeNet | 99.50 | |||
| [18] | VGG19 +CNN | GitHub+ cancer X-ray and CT images [28] |
98.05 | Imbalanced data |
| ResNet152V2 | 95.31 | |||
| ResNet152V2 + GRU | 96.09 | |||
| ResNet152V2+ Bi-GRU | 93.36 | |||
| [19] | ResNet50 | Cohen [25] Kaggle [29] |
93.01 | Imbalanced data Overfitting issue |
| ResNet101 | 97.22 | |||
| [20] | VGG-16 | Khan et al. [22] Ozturk [20] |
79.58 | Overfitting issue |
| [21] | FocusCOVID | Kaggle-1 [30] Kaggle-2 [31] | 95.20 | Cannot provide the optimal accuracy. |
| [22] | CoroNet | Chest X-ray Images [32] | 89.6 | This method is slow. |
| [23] | CheXImageNet | Cohen [25] Kaggle [29] |
100 | Overfitting issue |
| [24] | ResNet50 | Kaggle chest X-ray [33] RSNA pneumonia [34] |
91.13 | Cannot provide the optimal accuracy. |
| MobileNet | 93.73 | |||
| Hybrid model | 94.43 |