Table 10.

Performance analysis of various deep learning approaches in diagnosis of COVID-19.

Author	Month & Year	Method	Classes	Accuracy	Specificity	Sensitivity	AUC	Ref
Chen et al.	Jan, 2020	UNet++ (Retrospective testing per patient)	2	95.24%	93.55%	100%	—	[115]
Wang et al.	Jan, 2020	DenseNet121-FPN, COVID19-Net	2	85%	71.43%	79.35%		[75]
Li et al.	Mar, 2020	COVNet	3	—-	96%	90%	0.96	[116]
Asnaoui et al.	Mar, 2020	MobileNet_V2	2	96.27%	94.61%	98.02%	—-	[117]
		Inception_Resnet_V2		96.09%	93.88%	98.53%
		Resnet50		96.61%	94.92	98.43%
Abbas et al.	Mar, 2020	DeTrac-ResNet-18	3	95.12%	91.87%,	97.91%,	—-	[62]
Ozkaya et al.	Mar, 2020	CNN using Deep Features Fusion and Ranking Technique	—	98.27%	97.60%	98.93%	—-	[70]
Wang et al.	Mar, 2020	Tailored CNN	3	92.3%	—	—-	—-	[118]
Chowdhury et al.	Mar, 2020	Sgdm-SqueezeNet	3	98.3%	99.0%	—	—	[119]
Apostolopoulos et al.	Apr, 2020	CNN with transfer learning	3	96.78%	98.66%	96.46%	—	[120]
Afshar et al.	Apr, 2020	Capsule Network	4	95.7%	95.8%	90%	0.97	[121]
Butt et al.	Apr, 2020	Multiple CNN's	3	—-	92.2%	98.2%	0.99	[122]
Li et al.	Apr, 2020	DenseNet	3	88.9%	—	—-	—-	[123]
Ucar et al.	Apr, 2020	Bayes-SqueezeNet	3	98.3%	99.1%	—-	—-	[124]