Purchase this article with an account.
Michael Gale, Gareth Harman, Jimmy Chen, Mark E Pennesi; Repeatability of Flood-Illuminated Adaptive Optics Imaging in Subjects with Retinitis Pigmentosa. Invest. Ophthalmol. Vis. Sci. 2018;59(9):668.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine the intersession repeatability of cone density and spacing via flood-illuminated adaptive optics (AO) imaging in subjects with Retinits Pigmentosa (RP), in order to better differentiate between random change due to imaging variability vs pathology-driven loss of photoreceptors.
All testing was approved by OHSU IRB. We used the RTX1 flood-illuminated AO camera (Imagine Eyes: Orsay, France) to image 10 subjects with RP (22-57 years old; 2 F, 8 M). For each subject, 25 4°x4° images were acquired 3 times on the same day. Each set of 3 corresponding images were registered in i2K Retina. Cone photoreceptors were identified using a custom-built MATLAB cone counting algorithm. Nine equally spaced 100 x 100mu regions of interest (ROIs) were used to determine cone density variation across the 3 sessions, with the central ROI determined from an overlay of the 3 registered images. A subset of subjectively “poor” and “good” quality images were selected for further subgroup analysis by 3 independent graders, based on the ability to clearly visualize a hexagonal cone mosaic.
Image quality varied substantially across subjects. Average cone density was inversely correlated to the distance from the fovea, and coefficient of variation (CoV) ranged from 10.1-21.0% across the measured foveal eccentricities. In a prior study of normal subjects, CoV ranged from 0.7-7.7% in comparable sampling regions (Feng et al 2015). There was no correlation between cone density and CoV or repeatability. Repeat identification of cones was higher in good-quality images compared to poor-quality images (80.2% vs 41.5%). Cone spacing as determined by nearest neighbor distance was similar in both image sets, but standard deviation was increased in poor-quality images (1.79 vs 1.62).
Current flood-illuminated AO technology appears to be most effective when imaging RP subjects with higher quality images. Our findings indicate that average automated cone density can be repeatable in areas of poor image quality; however, the signals identified as cones are much more likely to be imaging noise rather than true cones based on reduced percent repeat identification of cones in poor images. When comparing quantitative cone density and cone spacing metrics across imaging sessions in RP patients, caution must be used as variability is higher than healthy subjects, and increases as image quality degrades.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only