Abstract
Purpose :
Successful trials for inherited retinal diseases must account for the interplay between retinal structure and function. This interplay, however, can be frustratingly complex; for example, in both Stargardt disease and age-related macular degeneration, steady rates of progressive retinal atrophy produce nonlinear and variable changes in best corrected visual acuity (BCVA). To address this challenge, we constructed an integrated computational model to explore, predict, and explain structure-function relationships for progressive retinal diseases involving atrophy.
Methods :
Our software extends previously published models of retinal atrophy in choroideremia and Stargardt disease to include BCVA and imaging predictions. The model is implemented in Julia, allowing for rapid prototyping, interaction, and statistical evaluation. The baseline model links atrophic changes in the RPE, under different scenarios for background risk (Figure A, B) and neighbor-induced cell death, to changes in preferred retinal locus and logMAR BCVA (Figure C, D). The quantitative output of the model can then be used to render hypothetical progression patterns in other imaging modalities.
Results :
In some patients with early-onset Stargardt disease, atrophic changes begin in a parafoveal ring, simultaneously extending toward and away from the macula. Once the fovea is lost, acuity drops suddenly (Figure F). Our model successfully recapitulates these changes. By overlaying the output of the model on control autofluorescence images (Figure E), we can visualize expected patterns of atrophy over time.
Conclusions :
Using a simplified model of atrophy, we can simulate the drop in visual acuity that accompanies parafoveal atrophy in our cohort of early Stargardt cases ("waterfall"; see Whitmore et al., ARVO abstract from 2023). Differences between observed outcomes and computational predictions suggest refinements for the model, such as incorporating prodromal stages and reduced functionality of surviving retina. Future work may also extend the model to include Goldmann visual field data.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.