TABLE 4. Summary of Deep Learning Based COVID-19 Diagnosis in X-Ray Images Using Customized Network.

Authors	Data Sources	Number of images	Number of classes	Partitioning	Techniques	Performances (%)
Wang et al. [99]	COVID-19 X-ray image database [88], RSNA Pneumonia Detection Challenge dataset [95]	13, 800	3 (COVID-19, non-COVID-19, normal)	Training=90%, Testing=10%	COVID-Net (CNN)	Accuracy= 92.4, Sensitivity=80, Precision=88.9
Ucar and Korkmaz [124]	COVID-19 X-ray image database [88], COVIDx Dataset [99], Kaggle chest X-ray pneumonia dataset [133]	2839 (COVID-19=45, normal=1203, pneumoniae] 591)	3 (COVID-19, normal, pneumonia)	Training=80%, Testing=10%, Validation=10%	Bayes-SqueezeNet	Accuracy=98.26, Specificity=99.13, Fl-Score=98.25, MCC=97.39, Correctness=98.26, Completeness=98.2 6
Khan et al. [125]	COVID-19 X-ray image database [88], Kaggle chest x-ray repository [93]	1251 (COVID-19=284, normal=310, pneumonia bacterial=330, pneumonia viral= 327)	4 (COVID-19, normal, pneumonia bacterial, pneumonia viral)	Training=80%, Validation=20%	CoroNet (CNN)	Accuracy=89.5, Sensitivity=100, Precision=97, Fl-Score= 98
Rahimzade h and Attar [126]	COVID-19 X-ray image database [88], RSNA Pneumonia Detection Challenge dataset [95]	15085 (covro-19=180, pneumonia= 6054, normal= 8851)	3 (COVID-19, pneumonia, normal)	5- fold cross-validation	Concatenated CNN	Accuracy=99.50, Sensitivity=80.53, Specificity=99.56, Precision=35.27
Mukheijee et al. [127]	covid-chestxray-dataset [88], Kaggle chest x-ray repository [93]	260 (COVID-19=130, non-COVID=130)	2 (COVID-19, non-COVID)	5- fold cross-validation	Shallow CNN	Accuracy= 96.92, Sensitivity= 94.20, Specificity=100, Precision=100, Fl-Score=97.01, AUC=99.22
Li et al. [128]	COVID-19 X-ray image database [88], Kaggle dataset [89], Kermany et al. [90]	2239 (COVID-19=239, pneumonia=1000, normal=1000)	3 (COVID-19, pneumonia, normal)	5-fold cross-validation	DCSL	Accuracy=97.01, Sensitivity=97.09, Precision=97, Fl-Score=96.98
Khobahi et al. [129]	COVID-19 X-ray image database [88], RSNA Pneumonia Detection Challenge dataset [95], COVIDx Dataset [99]	18,529 (COVID-19=99, non-COVID-pneumonia=9579, healthy=8851)	3 (COVID-19, non-COVID pneumonia, healthy)	Training=90%, Testing=10%	CoroNet (AutoEncoders)	Accuracy=93.50, Sensitivity=93.50, Precision=93.63, Fl-Score=93.51
Alqudah et al. [130]	COVID-19 X-ray image database [88]	71 (COVID-19=48, non-COVID-19=23)	2 (COVID-19, non-COVID-19)	Training=70%, Testing=30%	CNN, SVM, RF	Accuracy=95.2, Sensitivity=93.3, Specificity=100, Precision= 100
Farooq and Hafeez [131]	COVIDx Dataset [99]	13, 800	4 (COVID-19, normal, bacterial, viral)	Training=90%, Testing=10%	COVID-ResNet (CNN)	Accuracy= 96.23, Sensitivity=100, Precision=100, Fl-Score=100
Afshar et al. [132]	COVIDx Dataset [99]	13, 800	3 (COVID-19, normal, non-COVID-19)	Training=90%, Testing=10%	COVID-CAPS (Capsule Network)	Accuracy 95.7, Sensitivity=90, Precision=95.8, AUC=97