Understanding the changes of cone reflectance in adaptive optics flood illumination retinal images over 3 years

Abstract

Purpose : The aim of this work is to provide an insight into cone reflectance variability over timescales ranging from minutes to years using an AO retinal camera. In addition, we evaluated if the position of the light source in a commercial AO camera affects cone detection and measures of cone density.

Methods : The data set consisted of 61 images acquired with an AO retinal camera (rtx1, Imagine Eyes, France) at 2.5 degrees and 4 degrees temporal from the fovea of the right eye of a healthy subject (female, 27 years old). The time separation between the images ranged from 45 minutes to 3 years. Most of the images were acquired with two positions of the light source, either on the pupil centre (“on” illumination) and 2 mm temporal (“off” illumination) from the pupil centre.
The analysis of the images was performed with a novel semi-automated method for the exclusion of the blood vessel shadows and used entirely automated cone detection and segmentation. We compared the cone density as measured with on and off-axis illumination in 10 sampling windows (0.16mmx0.16mm) over time. We measured the intensity values of the cones in all the images, separating the results according to time ranges of minutes, hours, days, months and years.

Results : The reflectance of the individual cones showed variability increasing logarithmically with the acquisition time difference (Figure 1). The position of the light source did not affect the cone density evaluation across the images and it ranged between 28000 and 17000 cones/mm² going further from the fovea (Figure 2). The histograms of the cone reflectance were similar for both illumination series.

Conclusions : We developed a method for the automated analysis and monitoring of the same location in AO retinal images over time. In the evaluation of cone density, the position of the light source does not affect the results. The causes of cone reflectance variability detected with an AO flood camera are still unclear, but they present long-term variations that increase with time separation.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.

 

Figure 1. Changes in cone reflectance over time

 

Figure 2. Bland-Altman graph of the cone density measured with on-axis and off-axis illumination at different times on the selected windows at 4 degrees, left, and at 2.5 degrees, right