Reliability of cone counts using an adaptive optics retinal camera

Background: To assess the reproducibility and repeatability of cone imaging in healthy human eyes, using the RTx-1 Adaptive Optics Retinal Camera and its proprietary cone-counting software. Design: Single-centre, prospective study. Participants: Ten healthy adults. Methods: Macular cones were imaged. Intrasession repeatability was assessed by comparing 10 consecutive acquisitions obtained by the same operator from each subject. For the intersession study, each subject was imaged five consecutive days. Interoperator reproducibility was also evaluated by comparing the images obtained from 10 different subjects by two independent operators. Finally, intergrader agreement was evaluated by comparing the cone counts measured by two masked graders. Main Outcome Measures: Mean cone density (cells/mm(2)), spacing between cells (mu m) and percentage of cones with six neighbours calculated on Voronoi diagrams were measured. Correlation coefficients, intraclass correlation coefficients, and coefficients of variation were calculated. Results: Correlation coefficient and intraclass correlation coefficient were respectively 0.81 and 0.96 between operators, and 0.97 and 0.98 between the two graders. The intrasession and intersession coefficients of variation were under 7%. The percentage of cells with six neighbours and the spacing between cones varied in the same proportion (coefficients of variation ranged from 1.66 to 10.05%). Conclusions: Overall, the test-retest variability of RTx-1 and its software was good in normal human eyes. Further studies in the normal clinical setting are mandatory.