Table 4.

A summary of different prepossessing approaches employed in deep learning-based automated detection of COVID-19

Preprocessing	Approach	Method	Reference
Addressing data imbalance	Data augmentation	Rotation, Translation, Cropping, Flipping etc.	[30–40, 51, 55, 66, 79, 81, 88, 95]
	Class resampling	Undersampling	[113, 114]
		Oversampling	[37, 49, 93, 114]
	Loss function	Focal Loss[115]	[51]
		Weighted Loss[116]	[49, 52, 88]
Image segmentation	Segmentation model	VB-Net[97]	[52, 93, 96, 98, 100]
		U-Net[103]	[48, 51, 95]
		V-Net[99]	[54]
		BCDU-Net[102]	[53]
Image quality enhancement	Contrast enhancement	Histogram Equalizetion	[30, 35, 48, 49, 51, 80, 93]
	Brightness changing	Adding or subtracting every pixel by a constant value	[38, 51, 80]
	Noise removal	Perona-Malik filter [117]	[35]
		Adaptive Total Variation [118]	[49]
	Edge sharpening	Unsharp Masking	[35]