Abstract
Purpose :
Retinal ischemia is an important feature in diabetic retinopathy (DR). We propose a 3D quantitative analysis method based on optical coherence tomography angiography (OCTA) to measure ischemic volumes.
Methods :
We acquired 3×3-mm macular OCTA volumes from the eyes of 20 healthy participants and 179 patients with diabetes (with or without DR) using a spectral-domain OCTA system (Avanti RTVue-XR, Visionix/Optovue, Inc). Masked trained retina specialists graded disease severity in these eyes based on the Early Treatment of Diabetic Retinopathy Study scale using 7-field fundus photography. We defined referable DR (rDR) as level 35 or worse or any level with diabetic macular edema (DME); and vision-threatening DR (vtDR) as level 53 or worse or any level with DME. Each OCTA retinal volume was normalized to the average thickness of the healthy retina after foveal registration and Bruch membrane flattening. We then enhanced, binarized, and skeletonized the vasculature in each normalized volume. The theoretical oxygen tension map of each volume was calculated using an oxygen diffusion model with zero-order kinetics. For the study, we assumed equal oxygen tension within the vessels and homogenous consumption throughout the retina. The reference oxygen tension map was established by combining all the healthy volumes. We then extracted a 3D ischemia map for each volume by comparison with the reference map (Fig. 1).
Results :
The lower 3D-oxygen tension and higher ischemic volume percentage were strongly correlated with increased DR severity (Table 1). The diagnostic performance measured by area under receiver operating characteristic curve (AUC) were 0.99 for diabetes, 0.86 for DR, 0.85 for rDR, and 0.80 for vtDR diagnoses. In addition, no healthy eye had ischemic volume detected by this method.
Conclusions :
We developed a quantitative method to simulate 3D oxygen tension and ischemic volume using OCTA in diabetic eyes. It may provide a surrogate for in vivo evaluating oxygen status in DR.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.