Table 2.
Summary of DL-based methods for COVID-19 pneumonia classification. C.V refers to cross-validation, Tr refers to training, Val refers to the validation, AE refers to autoencoder. Sens refers to the sensitivity. Spec refers to the specificity. Acc refers to the accuracy
| Author | Modality | Dataset | 2D/3D | All data | All COVID-19 | Network and technique | C.V | Sens (%) | Spec (%) | Acc (%) |
|---|---|---|---|---|---|---|---|---|---|---|
| Narin et al. (2020) | CXR | COVID-19/normal | 2D | 3141 | 341 | 5 pretrained CNNs | 5 | / | / | 96.1 |
| COVID-19/pneumonia | 1834 | 99.5 | ||||||||
| COVID-19/bacterial | 3113 | 99.7 | ||||||||
| Wang et al. (2020d) | CT | COVID-19/pneumonia/normal | 3D | 1,266 | 924 | DNN | / | Tr: 78.93 Val: 80.39 | Tr: 89.93 Val: 81.16 | / |
| Song et al. (2020) | CT | COVID-19/normal/bacterial | 2D | 275 | 88 | DRE-Net | / | 93 | 96 | 99 |
| Khalifa et al. (2020) | CXR | Pneumonia dataset | 2D | 624 | 50 | GAN + TL | / | / | / | 99 |
| Apostolopoulos and Mpesiana (2020) | CXR | COVID-19/normal/bacterial | 2D | 1427 | 224 | MobileNet v2 | 10 | 98.66 | 96.46 | 94.72 |
| Misra et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 6008 | 184 | Three ResNet models | 5 | / | / | 93.9 |
| Khoshbakhtian et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 8850 | 498 | AE : COVIDomaly | 3 | / | / | 76.52 |
| Nour et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 2905 | 219 | CNN+k-NN+SVM | / | / | / | 98.70 |
| Aslan et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 2905 | 219 | ANN+AlexNet | / | / | / | 98.97 |
| Oh et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 502 | 180 | ResNet-18 | / | 76.90 | 100 | / |
| Hall et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 2905 | 219 | Ensemble:Resnet50 and VGG16 | 10 | / | / | 91.24 |
| Jaiswal et al. (2020) | CXR | COVID-19/normal | 2D | 2492 | 1262 | TL and DenseNet201 | / | / | / | 99.82 |
| Das et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | / | / | Xception | / | 97.09 | 97.29 | 97.40 |
| Ismael and Şengür (2020) | CXR | COVID-19/normal | 2D | 380 | 180 | 5 pretrained models + SVM | / | / | / | 94.7 |
| Gupta et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 2905 | 219 | Pretrained models | / | / | / | 99.08 |
| COVID/non-COVID | 99.53 | |||||||||
| Makris et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | / | / | 5 pretrained CNNs | / | / | / | 95 |
| Afshar et al. (2020) | CXR | Bacterial, Non-COVID Viral, COVID-19 | 2D | / | / | 5 COVID-CAPS | / | 90 | 95.8 | 95.7 |
| Minaee et al. (2020) | CXR | COVID-19/normal | 2D | 5000 | 184 | 5 TL+pretrained models | / | 100 | 98.3 | / |
| Maghdid et al. (2020) | CXR+CT | COVID-19/normal | 2D | 526 | 238 | TL+AlexNet model | / | 72 | 100 | 94.1 |
| Benbrahim et al. (2020) | CXR+CT | COVID-19/normal | 2D | 320 | 160 | TL+InceptionV3 and ResNet50 | / | 72 | 100 | 99.01 |
| Islam et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 4575 | 1525 | LSTM+CNN | / | 99.2 | 99.9 | 99.4 |
| Yang et al. (2020a) | CXR | COVID-19/pneumonia/normal | 2D | 4,448 | 2,479 | 3D Inception V1 | 10 | / | / | 95.78 |
| 101 | 52 | 98.08 | 91.30 | 93.3 | ||||||
| Zulkifley et al. (2020) | CXR | COVID-19/pneumonia/normal | 2D | 1343 | 446 | Conditional GAN : LightCovidNet | 5 | / | / | 97.28 |
| Heidari et al. (2020) | CXR | COVID-19/normal | 2D | Total: 8504 Training: 6899 | Total: 445 Training: 366 | TL VGG-16 model | / | 98.0 | 100 | 94.5 |
| Shalbaf et al. (2020) | CT | COVID-19/normal | 2D | 746 | 349 | TL+ Ensemble of 15 pretrained models | / | / | / | 85 |
| Goel et al. (2020) | CT | COVID-19/normal | 2D | 2482 | 1252 | AE+random forest | / | / | 98.77 | 97.87 |
| Hemdan et al. (2020) | CXR | COVID-19/normal | 3D | 50 | 25 | COVIDX-Net | / | 100 | 80 | / |