Table 1.
Summary of some representative works of the state of the art in comparison with our proposal. As shown, none of them work in lung chest segmentation of images from portable devices and are able to work with a significantly limited dataset.
| Author | Objective | Strategy | Image types |
|---|---|---|---|
| Hassanien et al. (2020) | COVID-19 detection | Thresholding Support Vector Machine | General purpose chest radiographs |
| Mei et al. (2020) | COVID-19 detection | Combination of two CNN and an SVM to generate a joint model | Computerized Tomography Images |
| Sethy and Behera (2020) | COVID-19 detection | Resnet50 to extract deep features and SVM for classification | General purpose chest radiographs |
| Apostolopoulos and Mpesiana (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | Transfer learning from generic domains | General purpose chest radiographs |
| Islam et al. (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | CNN + Long Short-Term Memory networks | General purpose chest radiographs |
| Rahimzadeh and Attar (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | Concatenation of two CNNs | General purpose chest radiographs |
| Loey et al. (2020) | COVID-19 classification (COVID-19, normal, bacterial pneumonia, and viral pneumonia) with a limited dataset | Generative Adversarial Networks and Transfer Learning | General purpose chest radiographs |
| Ucar and Korkmaz (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | Bayesian-optimized CNN | General purpose chest radiographs |
| Ozturk et al. (2020) | COVID-19 classification (No, COVID-19, Pneumonia) | You Only Look Once object detection with DarkNet CNN | General purpose chest radiographs |
| de Moura et al. (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | U-Net CNN | General purpose chest radiographs |
| de Moura et al. (2020) | COVID-19 classification (Normal, COVID-19, Pneumonia) | U-Net CNN | Chest radiographs exclusively from portable devices |
| Yan et al. (2020) | Segmentation of lung and COVID-19 regions in CT images | U-Net CNN | Computerized Tomography Images |
| Fan et al. (2020) | COVID-19 lung infection segmentation | Inf-Net CNN | Computerized Tomography Images |
| Chen et al. (2020) | Lung and COVID-19 infected regions segmentation in CT | Residual Attention U-Net | Computerized Tomography Images |
| Zhang et al. (2020) | COVID-19 detection | Anomaly detection with a Confidence-aware anomaly detection CNN | General purpose chest radiographs |
| Alom et al. (2020) | COVID-19 detection, lung segmentation and infected region localization | Inception Recurrent Residual Neural Network for detection and NABLA-N network for segmentation | General purpose chest radiographs |
| Ours | Lung region segmentation (Normal, COVID-19, Pneumonia) with a limited dataset and poor image quality | Two-stage transfer learning from MRI glioma segmentation to general purpose chest radiographs to portable device chest radiographs with U-Net CNN | Chest radiographs exclusively from portable devices |