Table 4.

Overview of papers using deep learning approaches with their working procedure and performance metrics for COVID-19 case detection.

Reference	Modality	Method	Remarks	Performance Metrics and Results
Wang et al. (2020) [24]	Chest X-Ray	COVID-Net	A deep learning-based model with a total of 16,756 X-Ray images with multiclass classification (three) and also proposed a dedicated dataset of COVID19 X-Ray images named COVIDx.	Accuracy;
				COVID-Net was able to achieve an accuracy of 92.40% for the classification of COVID19 positive cases.
				Sensitivity;
				COVID-Net has achieved decent sensitivity, which is 91.0% for COVID-19 cases.
				Positive predictive value;
				The positive predictive value of this approach is 98.9%.
Sethy et al. [25]	Chest X-Ray	ResNet50+SVM	The proposed model classified the characteristics obtained from various CNN (Convolutional Neural Network) models of the SVM (Support Vector Machine Classifier) using X-Ray images (25 COVID-19 positive and 25 Normal). The study claims that ResNet50 with the SVM classifier produces better results.	Accuracy;
				The authors claimed that the accuracy of their model is 95.38% for COVID-19 case detection.
				Sensitivity;
				97.29% sensitivity is achieved through this model.
Narin et al. [26]	Chest X-Ray	Deep CNN ResNet-50	This study used three different CNN models (ResNet50, InceptionV3, and InceptionResNetV2) using 50 open access COVID-19 X-Ray images from Joseph Cohen, and 50 typical images from a Kaggle repository. Besides, their used non-COVID images are images of children aged between 1 and 5 years.	Accuracy;
				They obtained an accuracy of 98% from their proposed model.
				Sensitivity;
				The claimed that recall or sensitivity of their model is 96%.
				Specificity;
				However, this method provides 100% specificity in detecting COVD-19 patients.
Ioannis et al. [27]	Chest X-Ray	VGG-19	In this study, 224 approved COVID-19, 700 cases of Pneumonia, and 504 normal radiology images were used. They performed on both binary and 3-class classification using a transfer learning method.	Accuracy;
				They achieved a performance accuracy of 98% for the binary class problem and 93% for the 3-class problem.
				Sensitivity;
				This study achieved 92% of sensitivity.
				Specificity:
				VGG-19 based approach obtained 98% of specificity.
Hemdan et al. [28]	Chest X-Ray	COVIDx-Net	This study deployed deep learning models to diagnose COVID-19 patients using chest X-rays. It proposed a COVIDx-Net model that included seven CNN models with 50 Chest X-Ray images (25 COVID19 positives, 25 normal).	Accuracy;
				The highest accuracy obtained among these seven CNN models is 90%.
				Precision;
				Similar to accuracy, among the seven CNN models, the highest precision achieved by this model is 100%.
				Sensitivity;
				Moreover, the highest sensitivity obtained among the models is also 100%.
Ying et al. [35]	Chest CT	DRE-Net	This approach used CT (777 COVID-19 positive, and 708 healthy) images with a deep model built into ResNet50 called DRE-Net.	Accuracy;
				DRE-Net obtained an accuracy of 86.00%.
				Sensitivity;
				This study claimed 96% of the sensitivity in COVID-19 detection.
				Precision;
				The precision value achieved by this model is 80%.
Wang et al. [36]	Chest CT	M-Inception	The authors used the modified Inception (M-Inception) deep model using CT images containing 195 COVID-19 positive images and 258 COVID-19 negative images.	Accuracy;
				The obtained accuracy of this M-inception model is 82.90%.
				Sensitivity;
				This study claimed that they achieved a sensitivity of 81%.
				Specificity;
				Moreover, this method provides specificity of 84%.
Zheng et al. [37]	Chest CT	UNet+3D Deep Network	This method proposed a three-dimensional Deep CNN model to detect COVID-19 from CT images. Their dataset contains 313 COVID-19 positive images and 229 non-COVID images.	Accuracy;
				The accuracy gained by this model is 90.80%.
				Sensitivity;
				This study obtained 90.70% of sensitivity.
				Specificity;
				This model achieved 91.10% of specificity in detecting COVID-19 positive cases from CT images.
Xu et al. [38]	Chest CT	ResNet	This study was performed in detecting COVID-19 positive cases using ResNet coupled with CT images. Their dataset contains the images of 224 Viral pneumonia and 175 healthy images	Accuracy;
				The average accuracy achieved by the model from the perspective of CT cases as a whole is 86.7%.
				Sensitivity;
				In detecting COVID-19 positive cases, this study reported 86.7% of sensitivity.
				Precision;
				The precision obtained by this model is 81.03%.
Tulin et al. [39]	Chest X-Ray	DarkCovidNet	This proposed model is based on the DarkNet method that is completely automated with an end-to-end structure without the need for manual feature extraction. They have used a total of 1125 images (125 COVID-19 positives, 500 Pneumonia images, and 500 NoFindings images) to experiment with their developed model.	Accuracy;
				This method obtained an accuracy of 98.08% and 87.02% for binary and three classes, respectively.
				Sensitivity;
				The sensitivity achieved by this study is 85.35% and 95.13% for binary and three classes, respectively.
				Specificity;
				Similarly, the specificity is also 92.18% and 95.3% for binary and 3-classes, respectively.
Our Proposed Framework	Chest X-Ray	Faster R–CNN	A deep learning model to detect COVID-19 cases from Chest X-Ray images using faster R–CNN models with ten folds cross-validation technique. A real-time assessment tool for COVID-19 positive case detection. The dataset contains 183 COVID-19 positive X-Ray images and 13617 non-COVID X-Ray images.	Accuracy;
				This proposed framework performed on binary classification and achieved a mean accuracy of 97.36%.
				Sensitivity;
				The mean sensitivity achieved by this model is 97.65%.
				Specificity;
				Also, for the specificity, the mean specificity obtained for 10 fold cross-validation method is 95.48%.