Table 6.
Benchmarking table showing a comparison of proposed and existing segmentation-based classification models.
| Author and Year | Segmentation | Dataset—Chest X-ray (COVID-19 Images + Other Images) |
Technique | Accuracy | Accuracy Improvement * |
AUC |
|---|---|---|---|---|---|---|
| Alom et al. 2020) [30] | NABLA-N network Accuracy—94.66 Dice—88.46 Jaccard—86.50 |
Kaggle (390 + 234) |
Inception Recurrent Residual Neural Network (IRRCNN) model | 3 class-87.26% | 10.19% | 0.93 |
| Wehbe et al. (2021) [31] | NA | Private (4253 + 14,778) |
Ensemble CNN | 2 class-83% | 14.45% | 0.9 |
| Oh et al. (2020) [87] | DenseNet103 Jaccard+95.5% |
Kaggle + GitHub (180 + 322) |
ResNet-18 | 4 class-88.9% | 8.55% | NA |
| Teixeira et al. (2021) [81] | UNet Dice+98.2% |
RYDLS-20-v2 (503 + 2175) |
Inception V3 | 3 class-88% (F1 score) | 9.45% | 0.9 |
| Keidar et al. (2021) [88] | NA | Private (1289 + 2427) |
Ensemble model | 2 class-90.3% | 7.15% | 0.96 |
| Fang et al. (2022) [55] | CLSeg Dice—94.09 |
COVIDGR 1.0 dataset (426 + 426) |
SC2Net (novel CNN) | 3 class-84.23% | 13.22% | 0.94 |
| Abdulah et al. (2021) [89] | Res-CR-Net Dice—98 Jaccard—98 |
Private (1435 + 3797) |
Ensemble CNN | 2 class-79% | 18.45% | 0.85 |
| Bhattacharyya et al. (2021) [90] | GAN network Accuracy—NA |
GitHub (342 + 687) |
VGG-19 + Random Forest | 3 class-96.6% | 0.85% | NA |
| Hertel et al. (2022) [91] | ResUnet Dice—95 |
COVIDx5 + MIDRC-RICORD-1C + BIMCV dataset (4013 + 12,837) |
Ensemble model | 2 class-91% 3 class-84% |
6.45% | 0.95 |
| Aslan et al. (2022) [92] | ANN based segmentation Accuracy—NA |
COVID-19 Radiography database (Kaggle) (219 + 2905) |
DensenNet201+SVM | 3 class-96.29% | 1.16% | 0.99 |
| Xu et al. (2021) [93] | ResUNet Jaccard—92.50 |
GitHub (433 + 6359) |
ResNet50 | 5 class-96.32% | 1.13% | NA |
| Proposed | UNet Accuracy—96.35 Dice—94.88 Jaccard—90.38 |
COVID-19 Radiography database (Kaggle) (3611 + 9849) |
Xception | 5 class-97.45% | - | 0.998 |
* Accuracy improvement with respect to proposed work.