Table 6.
Comparison of the proposed system with existing systems in terms of accuracy.
| AUTHOR | CLASSES | TYPE | MODEL | ACCURACY | PROS | CONS |
|---|---|---|---|---|---|---|
| Shibly et al. [44] | 2 Class: (COVID-19: 183, Healthy: 13617) |
Chest X-Ray | R–CNN | 97.36% | faster R–CNN 10-folds cross-validation |
Limited data Data-set |
| Wang et al. [47] | 2 Class (COVID-19: 313, Healthy: 229) |
Chest CT | DeCoVNet (UNet + 3D Deep Network) | 90.0% | light-weight 3D CNN weakly-supervised lesion localization for COVID Detection. |
Limited data |
| Ozturk et al. [18] | 2 Class (COVID-19: 125, Healthy: 500) |
Chest X-Ray | DarkCovidNet | 98.08% | The heatmaps produced by the model can be evaluated by an expert radiologist. High Binary Classification accuracy |
Limited data Relatively low Muti-class accuracy |
| 3 Class (COVID-19: 125, Healthy: 500, Pneumonia: 500) |
87.02% | |||||
| Apostolopoulos et al. [17] | 3 Class (COVID-19: 224, Healthy: 504, Pneumonia: 700) |
Chest X-Ray | VGG-19 | 93.48% | Multiple models used for testing Multiple datasets used for evaluation. |
Limited no of evaluation metrics |
| 3 Class (COVID-19: 224, Healthy: 504, Pneumonia: 700) |
MobileNet v2 | 92.85% | ||||
| Wang et al. [15] | 3 Class (COVID-19: 53, Healthy: 8066, Pneumonia: 5526) |
Chest X-Ray | COVID-Net | 93.3% | Low architectural complexity | Data-set imbalance |
| Law and Lin [45] | 3 Class (COVID-19: 1200, Healthy: 1341, Pneumonia: 1345) |
Chest X-Ray | VGG-16 | 94% | Multiple Models used. Improved Transfer Learning accuracy using data augmentation |
Cant generalize results of data augmentation |
| Cengil and Cinar [46] | 3 Class (COVID-19: 1525, Healthy: 1525, Pneumonia: 1525) |
Chest X-Ray | AlexNet + EfficientNet-b0 + NASNetLarge + Exception | 95.9% | 3 different datasets used i.e. robust Hybrid Model High Performance metrics |
High model complexity |
| Khan et al. [24] | 3 Class (COVID-19: 284, Healthy: 310, Pneumonia: 657) |
Chest X-Ray | Crornet (Xception) |
95% | 4-Class Classification results High Accuracy for COVID-19 class |
Limited data for COVID-19 Class |
| 4 Class (COVID-19: 284, Healthy: 310, Viral Pneumonia: 327, Bacterial Pneumonia: 330) |
89.6% | |||||
| Proposed |
3 Class (COVID-19: 1784, Healthy: 1755, Pneumonia: 1345) |
Chest X-Ray | COVDC-Net | 96.48% | Balanced Dataset 4-Class, 3-Class Classification High Performance Metrics Achieved |
Hybrid Methods are computationally expensive |
|
4 Class (COVID-19: 305, Healthy: 375, Viral Pneumonia: 379, Bacterial Pneumonia: 355) |
90.22% |