Table 3.
Overview of some COVID-19 detection approaches using deep learning. Trn train, Val validation, Tst test, TV train and validation, CV cross-validation, ACC accuracy, PRE precision, REC recall, SEN sensitivity, SPE specificity, F1-SCR F1 score, CM calculated from Confusion Matrix, AUC area under curve
| Sl. No. | Study | Method | Modality | Class | Dataset | Train-test split | Performance |
|---|---|---|---|---|---|---|---|
| 1 | LV et al. [48] | Casecade SEMENet | Chest X-ray | COVID-19, Pneumonia, Normal | [26, 108] | Trn: 6386 images Val: 456 images Tst: 456 images | ACC: 97.14% F1-SCR: 97% |
| 2 | Bassi et al. [32] | CheXNet [85] | Chest X-ray | COVID-19, Pneumonia, Normal | [26, 61, 62, 108] | Trn: 80% Val: 20% Tst: 180 images | ACC: 97.8% PRE: 98.3% REC: 98.3% |
| 3 | Yamac et al. [40] | CSEN (CheXNet [85]) | Chest X-ray | COVID-19, Viral, and Bacterial Pneumonia | [62, 63, 67, 72] | Stratified 5-fold CV | ACC: 95.9% SEN: 98.5% SPE: 95.7% |
| 4 | Zhang et al. [77] | COVID-DA (Domain Adaptation) | Chest X-ray | COVID-19, Pneumonia, Normal | [61, 65] | Trn: 10,718 images Tst: 945 images | AUC: 0.985 PRE: 98.15% REC: 88.33% F1-SCR: 92.98% |
| 5 | Goodwin et al. [88] | 12 Models Ensembled | Chest X-ray | COVID-19, Normal, Pneumonia | [26] | Trn: 80% Val: 10% Tst: 10% | ACC: 89.4% PRE: 53.3% REC: 80% F1-SCR: 64% |
| 6 | Misra et al. [37] | ResNet-18 | Chest X-ray | COVID-19, Normal, Pneumonia | [26, 65] | Trn: 90% Tst: 10% | ACC: 93.3% PRE: 94.4% REC: 100% |
| 7 | Abbas et al. [30] | DeTraC- ResNet18 | Chest X-ray | COVID-19, Normal, SARS | [26] | Trn: 70% Tst: 30% | ACC: 95.12% SEN: 97.91% SPE: 91.87% |
| 8 | Chowdhury et al. [31] | SqueezNet (Best Model) | Chest X-ray | COVID-19, Viral Pneumonia, Normal | [61] | 5-fold CV | ACC: 98.3% PRE: 100% REC: 96.7% F1-SCR: 100% |
| 9 | Farooq et al. [34]. | COVID-ResNet | Chest X-ray | COVID-19, Normal. Viral, and Bacterial Pneumonia | [39] | Trn: 13,675 images Tst: 300 images | ACC: 96.23% PRE: 100% REC: 100% F1-SCR: 100% |
| 10 | Alqudah et al. [78] | AOCT-Net | Chest X-ray | COVID-19, NonCovid-19 | [66] | 10-fold CV | ACC: 95.2% SEN: 93.3% SPE: 100% PRE: 100% |
| 11 | Hall et al. [33] | 3 Models Ensembled | Chest X-ray | COVID-19, Pneumonia | [26, 62, 67] | 10-fold CV | ACC: 91.24% SPE: 93.12% SEN: 78.79% |
| 12 | Hemdan et al. [109] | COVIDX-Net | Chest X-ray | COVID-19, Normal | [26] | Trn: 80% Tst: 20% | ACC: 90% PRE: 83% REC: 100% F1-SCR: 91% |
| 13 | Apostolo- poulos et al. [79] | VGG19 (Best Model) | Chest X-ray | COVID-19, Normal, Pneumonia | [26, 62, 65] | 10-fold CV | ACC: 93.48% SEN: 92.85% SPE: 98.75% |
| 14 | Apostolo- poulos et al. [55] | Mobile- Netv2.0 (from scratch) | Chest X-ray | COVID-19, nonCOVID-19 | [26, 62, 65, 67] | 10-fold CV | ACC: 99.18% SEN: 97.36% SPE: 99.42% |
| 15 | Karim et al. [35] | Deep COVID- Explainer | Chest X-ray | COVID-19, Normal. Viral, and Bacterial Pneumonia | [62, 65, 110] | 5-fold CV | ACC: 96.77% (CM) PRE: 90% REC: 83% |
| 16 | Majeed et al. [83] | CNNx | Chest X-ray | COVID-19, Normal. Viral, and Bacterial Pneumonia | [26, 62, 65, 108] | Trn: 5327 images Tst: 697 images | SEN: 93.15% SPE: 97.86% |
| 17 | Minaee et al. [36] | SqueezNet (Best Model) | Chest X-ray | COVID-19, nonCOVID-19 | [26, 111] | Trn: 2496 images Tst: 3040 images | ACC: 97.73% (CM) SEN: 97.50% SPE: 97.80% |
| 18 | Narin et al. [76] | ResNet50 (Best Model) | Chest X-ray | COVID-19, Normal | [26, 63] | 5-fold CV | ACC: 98% PRE: 100% REC: 96% SPE: 100% |
| 19 | Punn et al. [49] | NASNetLarge (Best Model) | Chest X-ray | COVID-19, Normal, Pneumonia | [26, 65] | Trn: 1266 images Val: 87 images Tst: 108 images | ACC: 96% PRE: 88% REC: 91% SPE: 94% |
| 20 | Ozturk et al. [92] | DarkCovidNet | Chest X-ray | COVID-19, Normal, Pneumonia | [26, 112] | 5-fold CV | ACC: 87.20% PRE: 89.96% REC: 92.18% |
| 21 | Sethy et al. [84] | ResNet50 +SVM | Chest X-ray | COVID-19, Normal | [26, 63] | Trn: 60% Val: 20% Tst: 20% | ACC: 95.38% F1-SCR: 95.52% MCC: 90.76% |
| 22 | Wang et al. [64] | COVIDNet | Chest X-ray | COVID-19, Normal, Pneumonia | [39] | Trn: 13,675 images Tst: 300 images | ACC: 93.30% PRE: 98.90% REC: 91% |
| 23 | Ucar et al. [38] | COVIDiag-nosisNet | Chest X-ray | COVID-19, Normal, Pneumonia | [26, 39, 108] | Trn: 80% Val: 10% Tst: 10% | ACC: 98.26% PRE: 98.26% REC: 98.26% SPE: 99.13% |
| 24 | Sun et al. [98] | AFS-DF (Deep-Forest) | Chest CT scan | COVID-19, Pneumonia | Not available | 5-fold CV | ACC: 91.79% SPE: 89.95% SEN: 93.05% AUC: 96.35% |
| 25 | Javaheri et al. [53] | COVID-CTNet | Chest CT scan | COVID-19, Pneumonia, Normal | Not available | Trn: 90% Val: 10% Tst: 20 cases | ACC: 90.00% SEN: 83.00% SPE: 92.85% |
| 26 | Kang et al. [54] | Multiview Representaion Learning | Chest CT scan | COVID-19, Pneumonia | Not available | Trn: 70% Tst: 30% | ACC: 95.5% SEN: 96.6% SPE: 93.2% |
| 27 | Donglin et al. [100] | UVHL (Hypergraph Learning) | Chest CT scan | COVID-19, Pneumonia | Not available | 10-fold CV | ACC: 89.79% SEN: 93.26% SPE: 84% PPV: 90.06% |
| 28 | Zhu et al. [52] | Joint regression and Classification | Chest CT scan | COVID-19, Severity estimation | Not available | 5-fold CV | ACC: 85.91% |
| 29 | Ouyang et al. [93] | Attention ResNet34 +Dual Sampling | Chest CT scan | COVID-19, Pneumonia | [65] | TV set: 2186 images Tst: 2796 images | AUC: 0.944 ACC: 87.5% SEN: 86.9% SPE: 90.1% F1-SCR: 82.0% |
| 30 | Chen et al. [81] | Residual Attention U-Net | Chest CT scan | COVID-19 (Segmentation) | [62] | 10-fold CV | ACC: 89% PRE: 95% DSC: 94% |
| 31 | He et al. [80] | DenseNet169 (Self-supervised Transfer Learning) | Chest CT scan | COVID-19, nonCOVID-19 | [68] | Trn: 60% Val: 15% Tst: 25% | ACC: 86% F1-SCR: 85% AUC: 94% |
| 32 | Maghdid et al. [82] | Modified AlexNet | Chest CT scan | COVID-19, Normal | [26, 73] | Trn: 50% Val: 50% Tst: 17 images | ACC: 94.1% SPE: 100% SEN: 90% |
| 33 | Maghdid et al. [82] | Modified AlexNet | Chest X-ray | COVID-19, Normal | [26, 73] | Trn: 50% Val: 50% Tst: 50 images | ACC: 94% SPE: 88% SEN: 100% |
| 34 | Butt et al. [56] | ResNet18 +Location Attention | Chest CT scan | COVID-19, Normal, Viral Pneumonia | Not available | TV set: 85.4% Tst: 14.6% | ACC: 86.7% PRE: 86.7% REC: 81.3% F1-SCR: 83.90% |
| 35 | Song et al. [94] | DRE-Net | Chest CT scan | COVID-19, Normal, Pneumonia | Not available | Trn: 60% Val: 10% Tst: 30% | ACC: 86% PRE: 79% REC: 96% F1-SCR: 97% |
| 36 | Zheng et al. [51] | DeCovNet | Chest CT scan | COVID-19 | Not available | Trn: 499 images Tst: 131 images | ACC: 90.10% SEN: 90.70% SPE: 91.10% |
| 37 | Barstugan et al. [50] | GLSZM+SVM (Best Model) | Chest CT scan | COVID-19, nonCOVID-19 | [62] | 10-fold CV | ACC: 98.71% SEN: 97.56% SPE: 99.68% PRE: 99.62% |
| 38 | Shi et al. [96] | iSARF (Random Forest) | Chest CT scan | COVID-19 Pneumonia | Not available | 5-fold CV | ACC: 87.9% SEN: 90.7% SPE: 83.3% |
| 39 | Gozes et al. [95] | 2D and 3D CNN (ResNet-50) | Chest CT scan | COVID-19 Normal | Not available | Trn: 50 patients Tst: 157 patients | SEN: 98.2% SPE: 92.2% AUC: 0.996 |