Finding the most discriminative slab between healthy and diabetic eyes with OCTA

Abstract

Purpose : There are many approaches to determining diabetic retinal changes. When using optical coherence tomography angiography (OCTA), observed changes to vessel density and perfusion density are common. This pilot study was conducted to explore the best en face slab to help differentiate between healthy and diseased eyes based on the detection of vascular differences.

Methods : Eyes from healthy controls and eyes with diabetic retinopathy were imaged with PLEX ^® Elite 9000 (ZEISS, Dublin, CA) swept-source OCT using the Angio 6x6 mm scans centered on the fovea. Slabs of various thicknesses were extracted. The standard layer thicknesses were varied by shifting the border between the superficial and deep layer in increments of 5 µm to create a set of 5 offset slabs (0, 5, 10, 15 and 20 µm thicker superficial and thinner deep). Vessel density and perfusion measurements were determined for the 9 sectors of the standard ETDRS grid. The mean differences for these parameters were determined for a group of healthy and disease eyes. Paired t-tests were performed on these differences and the area under receiver operator characteristic curve (AUC) values computed. To account for multiple comparisons, the significance level was set at 0.05/90 = 0.00056.

Results : Nine healthy and twelve eyes with non-proliferative diabetic retinopathy were evaluated. Statistically significant (p<0.00056) differences were found in the superficial perfusion and vessel density measurements between the healthy and diseased eyes, but only in the temporal sectors. Comparing the offset slabs, there did not appear to be an offset that was more effective than the default setting for the deep layer, although a thicker layer might improve discrimination in the superficial layer (see Table 1). The AUC curves provided similar findings (see Table 2).

Conclusions : There appear to be more statistically significant differences in the superficial vessel density and perfusion measurements than in the deep layer for healthy and diseased eyes, and this is not changed by altering the border between them. This study would benefit from additional eyes.

This is a 2021 Imaging in the Eye Conference abstract.

 

Table 1: Offset P-values for each sector of the ETDRS grid

 

Table 2: AUC values for each sector of the ETDRS grid