3-D tracing of capillaries through three retinal vascular plexuses toward their closest superficial artery and vein in OCT angiography volumes

Abstract

Purpose : In Optical Coherence Tomography Angiography (OCTA), analysis of vasculature at the capillary level is limited by shadowing from more superficial vasculature and discontinuities due to eye motion. We overcome these issues by applying recent post processing algorithms for 3-D motion correction and shadow artifact removal and present an approach to automatically trace capillary vasculature through the vessel network to the closest connected artery and vein. The goal of this fundamental study is to advance visualization and analysis tools for capillary vasculature especially in the less studied intermediate (ICP) and deep capillary plexuses (DCP).

Methods : OCT and OCTA data was motion compensated and merged using an algorithm based on orthogonally acquired volumes, OCTA data was compensated for shadow artifacts (Ploner et al. ARVO AM 2018, 2017). Large vessels in the superficial vascular plexus (SVP) were labeled as arteries or veins manually. From each voxel a path is computed to the artery and vein that can be reached with minimal cost. Path cost is accumulated over the traversed voxels in the shadow compensated OCTA volume: bright voxels (vasculature) are cheap, whereas dark voxels (background) are expensive, constraining the path to vasculature while being robust to noise.

Results : Figure 1 shows images of the vascular plexuses of a 28 y/o healthy subject. Figure 2 shows flythroughs through the OCTA volumes, and a B-scan extracted manually along a vessel path starting at a superficial artery, passing through the ICP, DCP, and again the ICP to a superficial vein.

Conclusions : Software motion correction and shadow artifact compensation allow detailed capillary visualization in three retinal vascular plexuses. Automatic tracing of the vessel network toward the closest superficial artery or vein can guide the reader and may aid in understanding capillary vasculature.

This is a 2020 ARVO Annual Meeting abstract.

 

Top row: 3x3 mm OCT fundus, segmentation lines of the ICP and DCP in the OCT(A) volume, A/V labels. Columns: SVP, ICP, between ICP and DCP, and DCP of the OCTA volume without (left) and with shadow compensation (center, right). Right column images colored green/brown in close proximity to arteries/veins.

 

Enface slices of the shadow compensated OCTA volume, B-scans extracted along the dotted curve.