Abstract
Purpose:
To investigate the reliability and repeatability of an image quality - grading system for cone photoreceptors imaged on an adaptive optics (AO) camera.
Methods:
50 healthy subjects, aged 20-35 years old, with no eye diseases were recruited prospectively. The left eyes of the participants were imaged using an AO camera (rtx1, Imagine Eyes, Orsay, France) at 5 degrees of retinal eccentricity, twice. The 2 images acquired on each subject were anonymized and randomized and the resulting 100 independent unpaired images were graded. A set of AO images validated by two independent experts in AO images and confirmed by a master grader was used as the basis of the 3-level grading system. Two masked independent observers (A and B) graded all the 100 images twice. Grading concordance was assessed by percentage of perfect agreement and kappa statistics.
Results:
Exact intra-grader agreement on the quality of images for A and B were 90% and 79% respectively, (unweighted k statistics 0.85 and 0.68 respectively, P’s<0.001). Exact inter-grader agreement between A and B’s, 1st grade was 73% and 2nd grade in 76% (unweighted κ statistics 0.59 and 0.63 respectively, P’s < 0.001).
Conclusions:
The high intra and inter - grader agreement of observer A and B’s 1st and 2nd grades from this study demonstrates a high degree of reproducibility of the grading scores. This AO image grading system can be used as a reliable indicator for assessing the quality of images acquired on this AO camera prior to assessing changes in cone counts.