Abstract
Purpose :
Describe the distribution of area of non-perfusion (NP) and neovascularization (NV) on ultra-widefield fluorescein angiography in patients with diabetic retinopathy (DR).
Methods :
This retrospective, cross-sectional study included all ultra-widefield fluorescein angiograms taken for DR at the Kellogg Eye Center after approval from the UM IRB from January 2009 to May 2018 that included 651 eyes. Exclusion criteria included previous panretinal photocoagulation and images with significant media opacity (e.g. vitreous hemorrhage or significant cataract). These images were manually segmented for retinal surface areas of NP and NV. The total area per patient were organized by frequency in a histogram, and which were tested against power law and exponential distributions.
Results :
Analysis of areas of NV across a population of 189 patients with proliferative DR demonstrates a power law distribution with an R2 fit of 0.9672 and shown by least squares method to be a superior fit to an exponential distribution with p < 0.05. Areas of NP over 794 patients with DR demonstrated a fit with an exponential distribution, with a superior fit compared to the power law distribution with p < 0.05. When the far periphery was excluded, the R2 fit for the exponential distribution was 0.9618.
Conclusions :
NV in patients with diabetes follows a power law distribution, and NP follows an exponential distribution. The difference in event distribution suggests that though the two phenomena are related, they are of fundamentally different. While exponential distributions are seen in many gradual, progressive phenomena in nature, the power law distribution tends to occur in constructs which follow the principle of self-organized criticality, suggesting that NV may propagate as a catastrophic local event in an unpredictable manner.
This is a 2021 ARVO Annual Meeting abstract.