Table 1.
Deep learning implementations in COVID-19 datasets.
| Ref. | Dataset | Methods Used | Evaluation Metrics | Research Challenges |
|---|---|---|---|---|
| Kermany et al. (2018) | Optical Coherence Tomography (OCT) image dataset | DL framework using transfer learning | Accuracy, Cross-Entropy Loss,True Positive Rate and False Positive Rate | Use of image dataset from varied sources to ensure generic usability of the proposed model not included in the study |
| Rajaraman, Candemir, Kim, Thoma, and Antani (2018) | Anteroposterior Chest X-ray dataset of children (1–5 years) from Guangzhou Women and Medical Center, China | Customized CNN model- VGG16 | Accuracy, AUC, Precision, Recall, Specificity, F-Score and MCC | Random Sampling Process Used. Reliability of the model in highly non-linear locality pertaining to predictions not considered in the study |
| Wang, Kang, et al. (2020) | CT Images from Xi’an Jiaotong University First Affiliated Hospital, Nanchang University First Hospital, Xi’an No.8 Hospital of Xi’an Medical College | AI and DL based Framework | Accuracy, Specificity and Sensitivity | Analysis of the relationship between Hierarchical features of CT images and genetic, epidemological information not included in the study |
| Shan+ et al. (2020) | Chest CT scan Image dataset from Shanghai Public Health Clinical Center | Human-In-the-Loop Strategy, DL Based Segmentation Network - VB Net | Dice Similarity Index, Pearson CorrelationCoefficient, Time of Manual Contouring, Segmentation Accuracy | Quantification of Imaging Metrics and establishing correlation between syndromes, epdemicology and treatment responses not included in the study |
| Ghoshal and Tucker (2020) | X-ray Images of Posterior-Anterior (PA) part of the lungs from Dr. Joseph Cohen's GitHub repository augmented with Chest X-ray Images from the Kaggle Dataset | Dropweights based Bayesian Convolutional Network | Prediction Uncertainty and Accuracy | Evaluation of the results with traditional state-of-the-art models not performed. Consideration of ‘Omics’ dataset not included for better insights on image markers |
| Apostolopoulos and Mpesiana (2020) | X-ray Image dataset from GitHub and Cohen, Image dataset from Radiological Society of North America (RSNA), Radiopaedia, and Italian Society of Medical and Interventional Radiology (SIRM) | Transfer Learning based on CNN | Accuracy, Sensitivity and Specificity | More in-depth analysis using larger datasets and development of models capable of distinguishing between COVID-19 and other viral infectious diseases |
| Huang, Han, et al. (2020) | CT Image Dataset from Tongji Hospital, Wuhan, China | Convolutional Neural Network (CNN) Architecture | Opacification Percentage | A structured reasoning on the impact of COVID-19 viruses on the opacities not included |
| Hemdan, Shouman, and Karar (2020) | X-ray Image Dataset from Dr. Joseph Cohen and Dr. Adrian Rosebrock | COVIDX-Net comprising of Deep CNN Models | Accuracy, Precision, Recall and F1-Score | More in-depth analysis on larger datasets missing |
| Narin, Kaya, and Pamuk (2020) | X-ray Images from Dr. Joseph Cohen GitHub Repository | Deep CNN Models - ResNet50, InceptionV3 and Inception-ResNetV2 | Accuracy, ROC and Confusion Matrices | Implementation of the CNN Models on larger datasets to enhance classification performance not considered |