Table 1.

Comparative perspective with related AI works for COVID-19 detection

Model	Dataset used	Method	Results	Shortcoming (s)
Wang et al. [15]	15000 chest radiography images of confirmed COVID-19 positive and negative cases	Deep convolutional neural network called COVIDNet	Accuracy-92.4$\%$, Sensitivity-80$\%$	Comparatively lower accuracy and sensitivity
Li et al. [16]	4356 Volumetric chest CT images that included community acquired pneumonia (CAP) and other non-pneumonia cases	3-Dimensional Convolutional ResNet-50 network, termed COVNetDeep	AUC-0.96	High computational cost and requirement of professionals to analyze the results
Gozes et al. [17]	CT images from 157 COVID affected patients	ResNet-50	AUC-0.996	Relatively small testing dataset
Xu et al. [18]	618 CT samples from COVID-19 patients (219), influenza-A infected (224), and healthy individuals (175)	Location attention network using ResNet-18	Accuracy-86.7$\%$	Lower accuracy
Ghoshal et al. [19]	5941 Chest radiography images samples from 4 classes: healthy, bacterial pneumonia, non-COVID-19 pneumonia	Drop-weights based Bayesian CNNs	Accuracy-89.92$\%$	Lower accuracy
Wang et al. [20]	1065 CT images (325 COVID, 740 Viral Pneumonia)	Modified inception transfer-learning model	Accuracy-79.3$\%$, Specificity-83$\%$, Sensitivity-67$\%$	Lower accuracy and imbalanced dataset
Fang et al. [21]	133 CT images of COVID-19 patients	Multilayer perceptron combined with an LSTM	AUC-0.954	Relatively smaller dataset size and lower accuracy
Jin Feng et al. [22]	970 CT images of COVID-19 positive and 1385 COVID-19 negative patients	2-Dimensional CNN	Accuracy-94.98$\%$, AUC-0.979	Lower accuracy and lack of generalization
Jin et al. [23]	1136 CT images (723 COVID-19 positive)	3-Dimensional UNet and ResNet-50	Specificity-92.2$\%$, Sensitivity-97.4$\%$	Lower accuracy
Narin et al. [24]	Chest X-ray images from 50 COVID-19 positive and 50 COVID-19 negative patients	ResNet-50	Accuracy-98$\%$	Relatively small testing dataset
Chowdhury et al. [25]	1341 Normal, 1345 Viral Pneumonia and 190 COVID-19 chest X-ray images	Combination of AlexNet, ResNet-18, DenseNet-201, and SqueezeNet	Accuracy-98.3$\%$	High computational cost, large number of training hyperparameters, and class imbalance
Maghdid et al. [26]	170 X-ray and 361 CT images	CNN augmented with a pre-trained AlexNet using transfer learning	Accuracy-98$\%$ for X-ray images, Accuracy-94.1$\%$ for CT images	High computational cost and lack of implementation in smart healthcare