Table 1.
Review of related works.
| Ref | Dataset | Modality | Number of cases | Pre-Processing | DNN | Post-Processing | Performance criteria |
|---|---|---|---|---|---|---|---|
| [54] | Clinical | CT | 3000 COVID-19 Images, 3000 Non-COVID-19 Images |
Patches Extraction |
VGG-16, GoogleNet, ResNet-50 |
Feature Fusion, Ranking Technique, SVM |
Acc 98.27 Sen 98.93 Spec 97.60 Prec 97.63 |
| [65] | Datasets from [66] & [67] |
CT | 460 COVID-19 Images, 397 Healthy Control (HC) Images |
Data Augmentation (DA) |
CNN Based on SqueezeNet |
Class Activation Mapping (CAM) |
Acc 85.03 Sen 87.55 Spec 81.95 Prec 85.01 |
| [68] | Various Datasets | CT | 2373 COVID-19 Images, 2890 Pneumonia Images, 3193 Tuberculosis Images, 3038 Healthy Images |
– | Ensemble DCCNs |
– | Acc 98.83 Sen 98.83 Spec 98.82 F1-Score 98.30 |
| [69] | Clinical | CT | 98 COVID-19 Patients, 103 Non-COVID-19 Patients |
Visual Inspection |
BigBiGAN | – | Sen 80 Spec 75 |
| [55] | Clinical | CT | 148 Images from 66 COVID-19 Patients, 148 Images from 66 HC Subjects |
Visual Inspection |
ResGNet-C | – | Acc 96.62 Sen 97.33 Spec 95.91 Prec 96.21 |
| [70] | COVID-CT Dataset |
CT | 349 COVID-19 Images, 397 Non-COVID-19 Images |
Scaling Process, DA |
Multiple Kernels-ELM -based DNN |
– | Acc 98.36 Sen 98.28 Spec 98.44 Prec 98.22 |
| [56] | Clinical | CT | 210,395 Images From 704 COVID-19 Patients and 498 Non-COVID-19 Subjects |
DA |
U-net Dual-Branch Combination Network |
Attention Maps | Acc 92.87 Sen 92.86 Spec 92.91 |
| [67] | Various Dataset | CT | 2933 COVID-19 Images | Deleting Outliers, Normalization, Resizing |
Ensemble DNN |
– | Acc 99.054 Sen 99.05 Spec 99.6 F1-Score 98.59 |
| [71] | Clinical | CT | 320 COVID-19 Images, 320 Healthy Control Images |
Histogram Stretching, Margin Crop, Resizing, Down Sampling |
FGCNet | Gradient- Weighted CAM (Grad-CAM) |
Acc 97.14 Sen 97.71 Spec 96.56 Prec 96.61 |
| [72] | Clinical | CT | 180 Viral Pneumonia, 94 COVID-19 Cases |
ROIs Extraction |
Modified Inception |
– | Acc 89.5 Sen 88 Spec 87 F1-Score 77 |
| [57] | Clinical | CT | 3389 COVID-19 Images, 1593 Non-COVID-19 Images |
Segmentation, Generating Lung Masks |
3D ResNet34 with Online Attention |
Grad-CAM | Acc 87.5 Sen 86.9 Spec 90.1 F1-Score 82.0 |
| [73] | COVIDx-CT Dataset |
CT | 104,009 Images From 1489 Patient Cases |
Automatic Cropping Algorithm, DA |
COVIDNet-CT | – | Acc 99.1 Sen 97.3 PPV 99.7 |
| [74] | Various Datasets | CT | 349 COVID-19 Images, 397 Non-COVID-19 Images |
Resizing, Normalization, Wavelet-Based DA |
ResNet18 | Localization of Abnormality |
Acc 99.4 Sen 100 Spec 98.6 |
| [58] | COVID-CT | CT | 345 COVID-19 Images, 397 Non-COVID-19 Images |
Resizing, DA | Conditional GAN ResNet50 |
– | Acc 82.91 Sen 77.66 Spec 87.62 |
| [75] | Clinical | CT | 151 COVID-19 Patient, 498 Non-COVID-19 Patient |
Resizing, Padding, DA |
3D-CNN | Interpretation by Two Radiologists |
AUC 70 |
| [59] | SARS-CoV-2 CT-Scan Dataset |
CT | 1252 CT COVID-19 Images, 1230 CT non-COVID-19 Images |
– | GAN with Whale Optimization Algorithm |
– | Acc 99.22 Sen 99.78 Spec 97.78 F1-score 98.79 |
| [66] | Various Datasets | CT | 1684 COVID-19 Patient, 1055 Pneumonia, 914 Normal Patients |
Resizing | Inception V1 | Interpretation by 6 Radiologists, t-SNE Method |
Acc 95.78 AUC 99.4 |
| [76] | Clinical | CT | 2267 COVID-19 CT Images, 1235 HC CT Images |
Compressing, Normalization, Cropping, Resizing |
ResNet50 | – | Acc 93 Sen 93 Spec 92 F1-Score 92 |
| [77] | Clinical | CT | 108 COVID-19 Patients, 86 Non-COVID-19 Patients |
Visual Inspection, Grey-Scaling, Resizing |
Various Networks |
– | Acc 99.51 Sen 100 Spec 99.02 |
| [60] | Various Datasets | CT | 413 COVID-19 Images, 439 Non-COVID-19 Images |
Feature Extraction with ResNet-50 |
3D-CNN | – | Acc 93.01 Sen 91.45 Spec 94.77 Prec 94.77 |
| [78] | Clinical | CT | 150 3D COVID-19 Chest CT, CAP and NP Patients (450 Patient Scans) |
Sliding Window, DA |
Multi-View U-Net 3D-CNN |
Weakly Supervised Lesions Localization, CAM |
Acc 90.6 Sen 83.3 Spec 95.6 Prec 74.1 |
| [61] | Various Datasets | CT | 449 COVID-19 Patients, 425 Normal, 98 Lung Cancer, 397 Different Kinds of Pathology |
Resizing, Intensity Normalization |
Autoencoder Based DNN |
– | Dice 88 Acc 94.67 Sen 96 Spec 92 |
| [79] | COVID-19 CT from [66] |
CT | 746 Images | – | GAN | – | Acc 84.9 Sen 85.33 Prec 85.33 |
| [80] | COVID-19 CT Datasets, Cohen |
CT | 345 COVID-19 CT Images, 375 Non-COVID-19 CT Image |
2D Redundant Discrete WT (RDWT) Method, Resizing |
ResNet50 | Grad-CAM, Occlusion Sensitivity Technique |
Acc 92.2 Sen 90.4 Spec 93.3 F1-Score 91.5 |
| [81] | SARS-CoV-2 CT Scan Dataset |
CT | 1262 COVID-19 Images, 1230 HC Images |
– | Convolutional Support Vector Machine (CSVM) |
– | Acc 94.03 Sen 96.09 Spec 92.01 Pre 92.19 |
| [82] | Chest CT and X-ray |
X-ray, CT |
5857 Chest X-rays, 767 Chest CTs |
– | Various Networks |
Heat Map | Acc 75 (CT) |
| [83] | medseg DlinRadiology |
CT | 10 Axial Volumetric CTS (Each Containing 100 Slices of COVID-19 Images) |
Resizing | VGG16, Resnet-50 U-net |
– | Acc 99.4 Spec 99.5 Sen 80.83 Dice 72.4 IOU 61.59 |
| [84] | BasrahDataset | CT | 50 Cases, 1425 Images | Gray-Scaling, Resizing |
VGG 16 | – | Acc 99 F1-Score 99 |
| [62] | Kaggle | CT | 1252 COVID-19 CT Images, 1240 non-COVID-19 CT Images |
Resizing, Normalization, DA |
Covid CT-net | heat map | Acc 95.78 Sen 96 Spec 95.56 |
| [85] | COVID-CT | CT | 708 CTs, 312 with COVID-19, 396 Non-COVID-19 |
Normalization | LeNet-5 | – | Acc 95.07 Sen 95.09 Prec 94.99 |