Table 1 Representative works for X-ray images based COVID-19 diagnosis according to [3]
References | Model of classification | Dataset |
|---|---|---|
Gaal et al. [14] | U-Net + adaptive histogram equalization with adversarial and contrast limits | 247 pictures obtained from the Japanese Society of Radiological Technology and 662 chest X-rays obtained from the Shenzhen dataset |
Abbas et al. [15] | Decompose, transfer, and compose CNN features of pre-trained models using ImageNet and ResNet + (DeTraC) | 80 typical CXR samples |
Narin et al. [16] | Transfer learning on a pre-trained ResNet50 model | Dr. Joseph Cohen’s public GitHub repository |
Wang et al. [17] | COVID-Net | 16,756 chest radiography pictures were collected from 13,645 patients |
Hemdanet al. [18] | COVIDX-Net | COVID-19 cases provided by Dr. Adrian Rosebrock |
Asnaoui et al. [20] | VGG16, VGG19, DenseNet201, Inception-ResNet-V2, InceptionV3, Resnet50, MobileNet-V2, and Xception have been fine tuned | 5856 pictures, 4273 of which are pneumonia and 1583 of which are normal |
Sethy et al. [13] | Deepfeatures fromResnet50 and SVM classification | – |
Ioannis [23] | Various fine-tune models: VGG19, MobileNet, Inception, Inception Resnet V2, Xception | 1427 X-ray images |
Ghoshal et al. [22] | Dropweights based Bayesian Convolutional Neural Networks | 5941 pictures of PA chest radiography divided into four groups Normal: 1583, Bacterial Pneumonia: 2786, Viral Pneumonia not caused by COVID-19: 1504, and COVID-19: 68 |
Farooq and Hafeez [21] | To boost model performance, they used a pre-trained ResNet50 architecture with the COVIDx dataset | COVIDx |