Comparison of two sampling methods for assessing the cone mosaic in adaptive optics retinal images

Abstract

Purpose : To evaluate two different strategies for defining sampling regions for analyzing the cone mosaic in images acquired with an adaptive optics (AO) camera in healthy and pathological eyes.

Methods : Thirty-four healthy subjects (27males and 7 females) and 2 patients with central areolar choroidal dystrophy (CACD) were examined. High resolution retinal images were acquired using AO camera (rtx1, Imagine eyes, Orsay, France), at eccentricities ranging from 0 to 8 degrees temporal. The AO images from each retina were stitched together using automated mosaic were then defined using two different protocols (modified methods reported by S.Feng et al. IOVS2015). The first method consisted in defining evenly distributed sampling windows along the central horizontal meridian, with a window size of 50x50 micron and a spacing of 1 degree (fixed-interval method). The second method was to select the 50x50 micron windows where cone density was maximal in the near vicinity of each fixed-interval windows (peak density method). Automated cone counting software (AO detect, Imagine eyes) was used to compute the detect cone positions in each sampling window. The results were manually corrected by three investigators. The results obtained from healthy subjects were compared with histological data (Curcio et al 1990) and a previous report. Both methods were also used to compare the cone densities of CACD patients with healthy control data.

Results : The mean age of subjects was 37 years (range24-55). The results obtained with the peak density method showed the best match with histological data. At eccentricities of less than 2 degrees, both methods underestimated cone density compared with histological data. Cone density in one of the CACD patient was found to be lower than density in control group at eccentricities from 1 to 6 degree horizontal. The other CACD patient showed differences of cone densities of more than 1 standard division at 3 to 4 degrees in comparison with control.

Conclusions : The peak density method provided the best agreement between clinical cone density results and histological data. The fixed-interval method was more affected by artifacts due to the shadowing effect of retinal vessels. As different sampling strategies can lead to significantly different density results, it is important to use a precisely defined sampling method when assessing the cone mosaic.

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