Abstract
Purpose :
For diabetic retinopathy (DR) patients, there is currently no published data comparing the irregularity of the ellipsoid zone inner border and its relationship to visual acuity. The purpose of this study was to evaluate the relationship between the two.
Methods :
To begin, a mathematical algorithm-based program was created to accurately capture the location of the pixels that correspond to the ellipsoid zone inner border in retinal optical coherence tomography (OCT) images. The variance of inner border pixel locations across the image was calculated to represent the overall irregularity. To eliminate varying size of serial OCT images taken over several visits contributing to falsely increased variance, each pixel location was normalized to the layer thickness. This was defined as the thickness between the ellipsoid zone inner border and outer border of the retina pigment epithelium (RPE)/Bruch’s membrane complex.
Patients’ visual acuity (VA) changes and their OCT images were then evaluated. Inclusion criteria included VA better than 20/100 in DR patients and availability of OCT images demonstrating stable or no inner retinal cysts. Patients were split into two groups; group 1 (G1) included those who had VA difference of ≥ 2 lines, and the rest in group 2 (G2).
Results :
From 2016 to 2018, 122 eyes of 61 diabetic retinopathy patients with available OCT images were selected. A total of 115 data sets were collected. Each dataset consisted of two VA measurements and corresponding OCT images. 48 of 115 (41.7%) datasets were in G1 and 67 were in G2 (VA difference of < 2 lines). Compared to G2, G1 showed a statistically significant correlation between worsening VA and increased variability of inner border of the ellipsoid zone (p < 0.0001). The mathematical algorithm predicted worsening VA based on OCT images in 85.4% of the G1 dataset (41 out of 48), while in the G2 dataset, the algorithm was 53.3% accurate.
Conclusions :
This variance of the inner border of the ellipsoid zone has previously not been identified as limiting visual acuity in DR patients thus highlighting this as an important structure to assess in these patients. Further, our mathematical algorithm predicts clinically significant visual acuity change ≥ 2 lines, introducing it as an additional tool to help determine visual prognosis in the clinic setting in DR patients.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.