The macula was scanned using a Zeiss Cirrus 5000 HD-OCT (Zeiss Meditec. Inc., Jena, Germany). This instrument uses a wavelength of 840 nm with a scanning rate of 68,000 A-scans per second. For all acquisitions, the FastTrac retinal-tracking technology was used to reduce motion artifacts. The resulting 6 × 6-mm angiography cube contains 245 B-scan slices. Each B-scan consists of 245 A-scans, and each A-scan is 1024 pixels deep. The 6 × 6-mm angiography images were obtained. Scanned images with a signal strength of less than 7 (from a maximum of 10) were not used in subsequent analysis. Moreover, the images that were decentered, poorly focused, or horizontally misaligned were excluded from the analysis.
We acquired segment slabs of the superficial (
Fig. 1a) and the deep (
Fig. 1b) vascular plexuses of the retina from the 6 × 6-mm angiography images. The outer border of the inner plexiform layer was used to produce each slab in the customized settings of OCT-A (
Figs. 1c–d). In cases with inaccurate segmentation, we manually adjusted the boundary between the superficial and deep vascular layers. The projection artifacts of the superficial layer were removed in the deep layer images by using a software built-in Zeiss OCT-A viewer. All images were exported into the ImageJ 1.50 software (National Institutes of Health, Bethesda, MD). FAZ was manually bordered by using the polygon selection tool (
Fig. 2a), after which its area and circularity (formula= 4 × π × area/[perimeter]
2) were automatically measured using ImageJ software.
23 A circularity value closer to 1 indicates a circular shape, and closer to 0 indicates more irregular shape.
24 We analyzed the vessel density and perfusion index of the 6 × 6-mm macula except the FAZ (central foveal 0.5-mm radius area). Vessel density (mm
−1) was calculated after skeletonization (
Fig. 2b) of the binarized image according to the methods described in previous studies.
25 The perfusion index was defined as spatial averaging output after image binarization (
Fig. 2c) by using autothresholding at 3 SDs above the mean noise level.
25 Only one eye of each patient was selected for the analysis (the eye with the better OCT-A image quality). Two independent readers (EYC and SEP) obtained and evaluated the OCT-A findings, and the average values were used for the statistical analysis.