Purpose
Retinal capillary system is comprised of four capillary networks. The purpose of this study was to develop a method to visualize each of the four capillary networks using optical coherence tomography (OCT) angiography technology.
Methods
Raster scans of 256A-Scan×256B-Scan by spectral-domain OCT (SD-OCT) were repeated four times within the 4.5 mm × 4.5 mm square macular region. Three-dimensional angiographic images were reconstructed by a motion-contrast method from the four data set. Automated layer boundary segmentation was performed on the intensity-based OCT images generated from the same data set. Angiographic images between the posterior boundary of and x-pixel (1 pixel= 4.2 μm) anterior line of the retinal nerve fiber layer (RNFL) were defined as RNFL capillary plexus (RNFLP). Angiographic images between the two lines y-pixel, anterior and posterior to the anterior boundary of the inner nuclear layer (INL) were defined as intermediate capillary plexus (ICP). Angiographic images between the two lines z-pixel, anterior and posterior to the posterior boundary of the INL were defined as deep capillary plexus (DCP). Angiographic images between RNFLP and ICP were defined as superficial capillary plexus (SCP). Visualization of capillary patterns were compared by changing the number of x, y, and z pixels in ten normal eyes.
Results
Images of the four capillary networks showed distinct capillary patterns. A complete image of RNFLP was obtained when x ≥ 3 pixels. Even if the number of x-pixels was increased, no capillary patterns were obtained in the parafoveal region and temporal raphe. SCP, ICP and DCP showed dense mesh patterns, which were different among the three networks. However, a partial overlap of capillary patterns of ICP was observed with that of SCP, and an overlap of some capillary patterns of SCP was observed with that of ICP. These partial overlaps changed with the number of y, but it did not disappear. In contrast, little overlap of the other plexus was found in the DCP. More than seven pixels in z were required to obtain a complete image of DCP.
Conclusions
Visualization of individual retinal capillary networks was possible based on segmentation of layer boundaries. No visualization of capillary patterns of RNFLP in the parafovea and temporal raphe and overlaps of partial capillary signals between SCP and ICP need to be addressed.