Table 8.
A comparison between the results of the proposed approach and the best results achieved by the other proposed approaches listed in [14], [15], [16], [17], [19].
| Approach | Number of original samples | Number of classes | Accuracy | Average precision |
Average Recall |
Average F score |
|---|---|---|---|---|---|---|
| MobileNet [14] | 224 COVID-19 504 Healthy 714 Pneumonia (400 bacterial ＋ 314 viral) |
2 classes | 96.78% | 96.46% | 98.66% | – |
| DarkCovidNet [15] | 125 COVID-19 500 No- Findings |
2 classes | 98.08%. | 98.03% | 95.13% | 96.51% |
| DarkCovidNet [15] | 125 COVID-19 500 No- Findings 500 Pneumonia |
3 classes | 87.02% | 89.96% | 85.35% | 87.37% |
| CNN-SA [16] | 403 COVID-19 721 Normal |
2 classes | 95% | 95% | 95% | 95% |
| CoroNet [17] | 284 COVID-19 310 Normal 330 Pneumonia Bacterial 327 Pneumonia Viral |
4 classes | 89.6% | 90% | 89.92% | 89.8% |
| CoroNet [17] | 284 COVID-19 310 Normal 657 Pneumonia (330 bacterial ＋ 327 viral) |
3 classes | 95% | 95% | 96.9% | 95.6% |
| CoroNet [17] | 284 COVID-19 310 Normal |
2 classes | 99% | 98.3% | 99.3% | 98.5% |
| Deep Bayes-SqueezeNet [19] | 76 COVID-19 4290 Pneumonia (bacterial ＋ viral) 1583 Normal |
3 classes | 98.3% | 98.3% | 98.3% | 98.3% |
|
Proposed GSA-DenseNet121-COVID-19 |
99 COVID-19 11 SARS 4 ARDS 6 Pneumocystis 2 Streptococcus 104 Healthy 80 pneumonia |
2 classes | 98.38% | 98.5% | 98.5% | 98% |