Abstract
Purpose :
The retinal vascular system in the macular region can be divided into three capillary plexuses, namely the superficial vascular plexus (SVP), intermediate (ICP) and deep capillary plexus (DCP). Being able to create an adequate OCTA visualization of these three plexuses could improve the understanding of various retinal diseases affecting the blood vessels. We therefore propose a fully automatic algorithm to compute shadow artifact compensated projections of the SVP, ICP and DCP.
Methods :
In order to compute the projection boundaries, the bottom of the Nerve Fiber Layer (NFL), Inner Plexiform Layer (IPL) and Outer Plexiform Layer (OPL) is first segmented with a multi-Dijkstra approach. This algorithm generates a graph from the B-Scans and computes the edge weights using the axial gradient. Improved reliability is achieved by computing the shortest paths for the three layer boundaries simultaneously with the spatial constraints such that the NFL/ IPL/ OPL is the top/ middle/ bottom boundary. This ensures that there are no discontinuities in the layer boundaries and that they have the correct spatial order. Shadow artifacts are compensated using an algorithm by Ploner et al (ARVO 2017). Finally the axial spreads of the plexuses are defined as [NFL-1; max(IPL-2, NFL-1)] for the SVP, [IPL-0.5;IPL+2] for the ICP and [mean(2/3*IPL+1/3*OPL)+2; OPL-3] and the associated enface projections are created using maximum projection.
Results :
Qualitative results for the segmentation and the enface projections are shown in Figure 1 and 2 respectively.
Conclusions :
Our fully automatic segmentation enables enface projections of individual capillary plexuses, opening up new research directions.
This is a 2020 ARVO Annual Meeting abstract.