Reliability of the cone counts by an Adaptive Optics Retinal Camera

Purpose To report the reproducibility and the repeatability of cone imaging in healthy human eyes, using the RTx-1™ Adaptive Optics Retinal Camera, (Imagine Eyes, Orsay, France) and AOdetect v0.1 cone counting software. Methods Cones were imaged in healthy emmetropic aduts. Mean cone density (cells/mm2), spacing between cells (µ) and percentage of hexagonal cells calculated on Voronoi diagrams, were measured with the proprietary software of the device, from a 0.3°square, positioned on a horizontal line crossing the fovea. The square was moved horizontally following a 10µ step, until reaching the parafoveal density peak. Intersession reproducibility was assessed by imaging 3 subjects, 5 consecutive day. Intrassession repeatability was assessed by comparing 10 acquisitions of 3 subjects, in a row. Inter-operator reproducibility was evaluated by comparing 10 subjects imaged by 2 operators. Inter grader reproducibiliy was evaluated by assessing 60 different images by 2 masked graders. Correlation coefficient and coefficients of variation(COVs) were calculated. Results The mean cone counts was respectively 23623±3831/mm2 and 22851±3543/mm2 for the first and the second operator (t test, p <0.05). Coefficient correlation was 0.81 between operators, and 0.97 between the 2 graders. The intrasession and intersession COVs were under 7%. The percentage of hexagonal cells and the spacing, varied in the same proportion (COV range from 1.66 to 10.05%). Conclusion Cone counts using the RTx-1™ device and its software are reliable in normal human eyes. Further studies in the normal clinical setting are needed.