A Novel Imaging Analysis Method to Quantify Fleck Area in ABCA4 Retinopathy

Abstract

Purpose : Stargardt disease is the most common inherited macular dystrophy and is often characterized clinically, by the presence of retina flecks. On autofluourescence imaging, a subset of these flecks are hyperfluorescent . The purpose of this study is to quantify hyperfluorescent flecks in a cohort of patients with ABCA4 retinopathy.

Methods : Patients with ABCA4 retinopathy were evaluated as part of a natural history study at the National Eye Institute and evaluated using Optos ultra-wide fundus autofluorescence imaging. Interactive machine learning software was utilized to segment hyperfluorescent flecks and quantify total area based on patient’s individual disc area using Image J software. A segmental ETDRS-like grid was utilized to subdivide the retina for further fleck quantification. Analysis was performed on demographic data, visual acuity and imaging data between eyes of individual patients and across the cohort using basic descriptive statistics and correlation with Spearman R.

Results : Thirty-eight patients (75 eyes) had identifiable hyperfluorescent flecks that could be segmented and were included in this analysis. Participant age ranged from 12 to 67 years (mean= 40 years) and included 23 females (61%) and 15 males (39%). Best corrected visual acuity (BCVA) ranged from 20/16 to 20/500 with a mean of 20/100. There was a strong correlation of BCVA between eyes for each patient (R = 0.88, p < 0.0001). Fleck area across all eyes ranged from 0.004 to 27.929 disc area. There was a statistically significant correlation between eyes for total fleck area (R = 0.84, p < 0.0001) as well as for all the rings and wedges of the adapted ETDRS grid. There was no correlation between fleck area and age or BCVA.

Conclusions : The ability to segment flecks through interactive machine learning software is an efficient and novel technique of quantifying and analyzing hyperfluorescent fleck lesions in patients with ABCA4 retinopathy. This imaging analysis method may aid in the monitoring of disease progression of retinal changes in patients with Stargardt disease, as well as serve as a possible outcome measure in future clinical trials.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.