Table 4.
Classification results.
| (A) Original AlexNet (n = 40) | ||||
|---|---|---|---|---|
| Classification | Accuracy | Specificity | Sensitivity | F-score |
| 1 COVID-19 | 96.864 ± 1.655% | 98.244 ± 1.713% | 91.656 ± 6.102% | 94.704 ± 3.078% |
| 2 Viral pneumonia | 96.272 ± 1.835% | 92.110 ± 4.613% | 97.833 ± 2.290% | 94.803 ± 2.389% |
| 3 Normal | 98.528 ± 0.675% | 98.145 ± 1.767% | 97.611 ± 1.978% | 97.854 ± 0.996% |
| Total | 95.832 ± 1.895% | 96.167 ± 1.540% | 95.700 ± 2.007% | 95.787 ± 1.953% |
| (B) Improved AlexNet (n = 40) | ||||
|---|---|---|---|---|
| Classification | Accuracy | Specificity | Sensitivity | F-score |
| 1 COVID-19 | 97.897 ± 0.941% | 98.980 ± 0.414% | 94.233 ± 3.325% | 96.516 ± 1.622% |
| 2 Viral pneumonia | 97.648 ± 1.000% | 94.656 ± 2.719% | 98.833 ± 0.960% | 96.676 ± 1.362% |
| 3 Normal | 98.872 ± 0.229% | 98.466 ± 1.208% | 98.278 ± 1.155% | 98.360 ± 0.329% |
| Total | 97.208 ± 0.955% | 97.367 ± 0.821% | 97.115 ± 1.027% | 97.184 ± 0.977% |
| (C) Improved AlexNet + SVM (n = 40) | ||||
|---|---|---|---|---|
| Classification | Accuracy | Specificity | Sensitivity | F-score |
| 1 COVID-19 | 98.834 ± 0.436% | 98.279 ± 0.555% | 97.975 ± 1.262% | 98.123 ± 0.712% |
| 2 Viral pneumonia | 98.642 ± 0.343% | 97.289 ± 1.350% | 98.833 ± 0.665% | 98.047 ± 0.478% |
| 3 Normal | 98.815 ± 0.217% | 98.934 ± 0.546% | 97.611 ± 0.644% | 98.266 ± 0.320% |
| Total | 98.145 ± 0.404% | 98.168 ± 0.384% | 98.140 ± 0.426% | 98.145 ± 0.410% |
| (D) The proposed model (Improved AlexNet + ReliefF + SVM) (n = 40) | ||||
|---|---|---|---|---|
| Classification | Accuracy | Specificity | Sensitivity | F-score |
| 1 COVID-19 | 99.082 ± 0.335% | 98.412 ± 0.577% | 98.650 ± 0.950% | 98.528 ± 0.540% |
| 2 Viral pneumonia | 99.082 ± 0.369% | 98.302 ± 1.061% | 99.056 ± 0.375% | 98.674 ± 0.528% |
| 3 Normal | 99.120 ± 0.273% | 99.218 ± 0.467% | 98.222 ± 0.861% | 98.715 ± 0.403% |
| Total | 98.642 ± 0.398% | 98.644 ± 0.388% | 98.643 ± 0.406% | 98.639 ± 0.398% |