Pseudoaveraging for denoising of OCT angiography: a deep learning approach for image quality enhancement in healthy and diabetic eyes

Table 2 Validation dataset grading results

	Grading Metrics	PseudoAvg vs. original (better ranked scan)		PseudoAvg vs. averaged (better ranked scan)		Averaged vs. Original (better ranked scan)
Scan Quality Metrics	Overall Scan Enhancement	PseudoAvg	p < 0.0001	PseudoAvg	p < 0.0001	Equal	p = 0.4575
	Background Noise Reduction	PseudoAvg	p < 0.0001	PseudoAvg	p < 0.0001	Equal	p = 1.0
	Motion (line) Artifact Reduction	PseudoAvg	p = 0.031	Averaged	p < 0.0001	Averaged	p < 0.0001
	Small Vessels Appearance	PseudoAvg	p < 0.0001	PseudoAvg	p < 0.0001	Original	p < 0.0001
	Continuity of FAZ Contour	Equal	p = 0.625	PseudoAvg	p = 0.0013	Original	p = 0.0011
Pathologic features Metrics	Appearance of Non-perfused Areas	PseudoAvg	p = 0.032	PseudoAvg	p = 0.0098	Equal	p = 0.5058
	Appearance of Pruned Vessels	PseudoAvg	p = 0.001	PseudoAvg	p = 0.0078	Equal	p = 0.1250
	Visibility of MAs	PseudoAvg	p = 0.0005	PseudoAvg	p = 0.0002	Equal	p = 0.0625

Comparisons and p-values using the Wilcoxon Signed Rank Test; PseudoAvg: Pseudoaveraged scan; FAZ: Foveal Avascular Zone; MA: microaneurysms

ISSN: 2056-9920