Table 1.
Representative works for Chest x-ray images based on the detection of COVID-19 infection.
| Literature | Models | Dataset | Performance |
|---|---|---|---|
| L. Wang et al. [12] | COVID-Net pre-trained with ImageNet | 5941 chest x-ray images across 2839 patient (1203 normal + 45 COVID19 + 660 non-COVID viral pneumonia + 931 bacterial pneumonia) | Accuracy of 92.4% for 2-classes and 83.5% for 4-classes |
| Hemdan et al. [13] | COVIDX-Net : based on DenseNet201, Inception v3, VGG19, MobileNet v2, Xception, InceptionResNet v2 and ResNet v2 | 50 X-ray images comprising 25 cases with COVID-19 and 25 cases without any infections | F1-scores of 0.89 for normal and 0.91 for COVID-19 |
| P. Kumar et al. [14] | Deep features from Resnet50 + SVM classifier | Dataset collected from GitHub and Kaggle comprising 25 cases with COVID-19 and 25 cases without any infections | Accuracy of 95.38% |
| Ozturk et al. [15] | DarkCovidNet | X-ray images comprising 125 with COVID-19, 500 normal and 500 Pneumonia cases | Accuracy of 87.02% for 3-class cases |
| Ioannis et al. [16] | VGG-19 | 1427 X-ray images including 504 images of normal cases, 700 images with confirmed bacterial pneumonia and 224 images with confirmed Covid-19 cases. | Accuracy of 93.48% for three classes. |
| Khan et al. [17] | CoroNet | Images collected from Kaggle repository, comprising 290 COVID-19, 1203 normal, 931 viral Pneumonia and 660 bacterial Pneumonia chest x-ray images. | Accuracy of 89.6% and 95% for 4 and 3 classes, respectively. |
| X. Xu et al. [18] | ResNet +Location Attention | 618 pulmonary CT samples (i.e., 175 healthy persons, 224 patients with Influenza-A, and 219 patients with COVID-19) | Accuracy of 86.7% |
| S. Wang et al. [19] | M-Inception | 99 Chest CT images (i.e., 55 viral pneumonia and 44 COVID-19) | Accuracy of 73.1%, along with a sensitivity of 74.0% and a specificity of 67.0% |
| L. Li et al. [20] | COVNet | 4356 chest CT images (i.e., 1735 pneumonia, 1325 non-pneumonia and 1296 COVID-19). | Specificity of 96%, sensitivity of 90%, and AUC of 0.96 |
| Y. Song et al. [21] | DeepPneumonia | Chest CT scans of 275 patients (88 patients infected with COVID-19, 101 patients infected with bacterial pneumonia, and 86 healthy) | Accuracy of 86.0% for (COVID-19 vs. bacterial pneumonia) classification and 94.0% for (COVID-19 vs. healthy) classification |
| B. Ghoshal et al. [22]. | Dropweights based Bayesian Convolutional Neural Networks | 5941 chest x-ray images across four classes (Bacterial Pneumonia: 2786, Normal: 1583, COVID-19: 68 and non-COVID-19 Viral Pneumonia: 1504). | Accuracy of 92,90% |
| J. Zhang et al. [23] | Deep CNN based on Backbone network | X-ray images from 1008 non-COVID-19 pneumonia patients and 70 COVID-19 patients | Sensitivity of 96.0% and specificity of 70.7% along with an AUC of 95.2%. |