Grading protocol

The scans were checked for segmentation errors and subsequently, the en-face OCTA images of the SCP and DCP were exported and masked. Then they were qualitatively and quantitatively evaluated for continuity of the vessels and presence of artefacts by 2 independent, experienced retinal imaging experts (S.Y. and M.R.M) according to a pre-specified grading protocol.[11] Briefly, the grading scale included 5 parameters: motion and image artefacts (1 = no artefacts, 0 = 1–4 artefacts, -1 = ≥ 5 artefacts or severe artefacts making reasonable evaluation of the microvasculature impossible), distinguishability of FAZ border (1 = FAZ border well distinguishable, 0 = middle, -1 = FAZ border barely/not distinguishable), and vessel continuity and discriminability of large and small vessels (1 = vessel continuity clearly preserved, 0 = middle, -1 = vessel continuity not preserved). Vessel continuity and discriminability of large vessels were only graded in SCP images.

As described in previous studies, motion artifacts were considered as present when there were doubling or displacement of retinal vessels, or characteristic white-line with corresponding discontinuity on the en-face image in B-scan direction, or continuous dark bands of varying width due to blink. Image artefacts were defined as any anomaly in addition to motion artefacts, such as segmentation artefacts, which leads to deviation of the slab; projection artefacts, which were from more superficial vasculature projecting on deeper structures; masking artefacts, which were low signal region blocked by the more anterior dense; loss of focus, ect.[5–7,12–15] Additionally, the number of clearly identifiable bifurcations in the SCP en-face images was counted to evaluate the vessel visibility. The counting procedure was as follows: a main, large vessel branch at 12 o`clock was chosen, and the number of identifiable, subsequent bifurcations towards the terminal capillary end were counted on the respective branch.[11]

After concordance analyses and the evaluation of the inter-grader reliability, a consensus grading was performed, including an individual score for each feature of each image.