Table 6.
Review of the literature for COVID19 detection using CXRs.
| References | Highlights | Pretraining | Dataset |
|---|---|---|---|
| Rajagopal (2021) | Combined deep learning and ML classifier | Yes | PedPneumonia, COVID-CXR, https://github.com/agchung |
| Jin et al. (2021) | Used deep feature followed by feature selection with SVM | Yes | PedPneumonia, COVID-CXR |
| Chowdhury et al. (2020) | Used deep ensemble feature generation | Yes | Mutiple datasets with different disorders |
| Khan et al. (2020) | XceptionNet based end-to-end training | Yes | PedPneumonia, COVID-CXR, COVIDDGR |
| Islam et al. (2020) | Used a combination of LSTM-CNN-based architecture | Yes | Combination of |
| publicly available data Pham (2021) | Used a multi-level classification approach for two and three disease classes | Yes | COVID-CXR, PedPneumonia, COVID-19 (kaggle), ActualMed (github) |
| Al-Rakhami et al. (2021) | Approach combines CNNs with sequential deep model | Yes | Data collected from various available sources |
| Duran-Lopez et al. (2020) | Proposed COVID-XNet, a custom deep learning model for binary classification | Yes | BIMVC, COVID-CXR |
| Gupta et al. (2021) | Proposed InstaCovNet-19, with ensemble generated from deep features | Yes | Chowdhury et al. (2020), COVID-CXR |
| Abbas (2021) | Class decomposition into sub-classes with pre-trained models | Yes | JSRT, COVID-CXR |
| Gour and Jain (2020) | Submodule stacking from pretrained and customized deep models | Yes | COVID-CXR, ActualMed, PedPneumonia |
| Malhotra et al. (2022) | Multi-task approach with segmentation, disease classification and | Yes | CheXpert, Chestxray14, BIMVC-COVID19, Various online sources |
| Pereira et al. (2020) | Feature ensemble of handcrafted and deep features | Yes | COVID-CXR, Chestxray14, Radiopedia Encyclopedia |
| Rahman T. et al. (2021) | Employed and compared different enhancement technique for performance improvement | Yes | PedPneumonia, BIMCV+COVID19 |
| Li et al. (2020) | On-device detection approach for CXR snapshots | Yes | PedPneumonia, COVID-CXR |
| Ucar and Korkmaz (2020) | Used Bayesian optimization with deep models for differentiating Pneumonia | Yes | PedPneumonia, COVID-CXR |
| Shi et al. (2021) | Knowledge transfer in the form of attention from teacher to student network | No | COVID-CXR, SIRM |
| Saha et al. (2021) | Used deep features with different ML classifiers | Yes | COVID-CXR, SIRM, PedPneumonia, Chestxray14, |
| Mahmud et al. (2020) | Used feature stacking generated from different resolutions | Yes | PedPneumonia, private |
Pretraining (yes/no) refers to use of weights of deep model trained on ImageNet dataset. Private refers that the data used is in-house and is not released publicly.