Purchase this article with an account.
Grace K. Han, Robert F. Cooper, Jessica C. Hsu, Min Chen, Joseph Carroll, Jessica Ijams Wolfing Morgan; Counting Cones in the Presence of Rods: A Reliability Study. Invest. Ophthalmol. Vis. Sci. 2016;57(12):4640.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Adaptive optics scanning light ophthalmoscopy (AOSLO) allows an in vivo assay of the arrangement of the photoreceptor mosaic. Recent advances in AOSLO imaging technology have enabled the visualization of the rod outer segments (OS), as well as the cone inner segments (IS). Understanding the reliability of these measurements is essential to quantifying longitudinal photoreceptor change. Here we examine the inter-observer reliability of identifying cone IS and cone OS in the presence of rod OS.
Five subjects (9 eyes) with no known pathology were imaged along the horizontal and vertical meridians using a custom AOSLO with both confocal and split detection modalities to visualize the cone and rod OS reflectance and cone IS mosaics, respectively. AOSLO images were semi-automatically montaged using a custom MATLAB algorithm. Regions of interest (ROIs) were selected from the confocal montage at 190 μm, and from both split detection and confocal montages at 900 and 1800 μm from the fovea. Due to fixational error, one subject’s montage did not extend to 1800 μm in one meridian, resulting in a total of 178 ROIs. Four observers, ranging from completely naïve to expert, manually identified cone OS and IS using custom software. Cone density was extracted from each ROI, and intraclass correlation coefficients with 95% confidence interval (CI) were calculated to assess the inter-observer agreement.
Inter-observer agreement was high in cone-only images (OS 190μm: 0.848, CI 0.764-0.911; IS 900μm: 0.908, CI 0.853-0.948; IS 1800μm: 0.889, CI 0.824-0.935), moderate in OS images at 900μm (0.683, CI 0.545-0.802) and poor in OS images at 1800μm (0.244, IC 0.085-0.440). Excluding the naïve observer’s cone density measurements substantially increased agreement within OS images (190μm: 0.991, CI 0.984-0.995; 900μm: 0.795, CI 0.678-0.881; 1800μm: 0.683, IC 0.521-0.811).
Inter-observer measurements of cone density, including those of expert observers, are more reliable in the absence of rod OS. This study underscores the need for additional reliability studies in eyes containing pathology where image quality can be substantially worse. Moreover, it is essential that observers are trained on cone selection, particularly when identifying photoreceptor OS in confocal AOSLO images.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only