Purchase this article with an account.
Suman Adhikari, Gwen Musial, Hanieh Mirhajianmoghadam, Alexander W Schill, Hope M Queener, Joseph Carroll, Jason Porter; Evaluating different methods for marking cones in simulated and in vivo retinal images. Invest. Ophthalmol. Vis. Sci. 2019;60(9):4593.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The use of different quantification methods for calculating cone packing metrics in living eyes has limited (1) understanding of the accuracy and variability of these measurements and (2) comparisons between studies. We sought to compare the performance of different cone marking methods within a given retinal area, or region of interest (ROI).
10 ROIs (37µm × 37µm and 100µm × 100µm) of random orientation were randomly extracted from simulated cone mosaics with uniform densities varying from 10,000-200,000 cones/mm2 (10,000 cones/mm2 increments) and placed with random orientations on real cone mosaics from 5 normal subjects acquired using adaptive optics scanning laser ophthalmoscopy at 0.3, 0.6, 0.9, 1.2, and 2.4mm eccentricities. We marked all cones fully within the ROI and partially within the ROI along (1) all 4 borders, (2) top and right borders only, and (3) bottom and left borders only. A custom program (Mosaic Analytics) calculated bound and unbound densities for all ROI sizes and marking methods. ANOVA or Kruskal-Wallis tests were used to compare density values between methods.
Marking all cones fully and partially contained within the ROI (Method 1) and computing unbound density yielded values that were significantly higher than simulated densities (mean difference ± SD = 11.3 ± 4.3%). Values were closest to simulated densities when performing Methods 1-3 and calculating bound density (mean % difference = 0.6 ± 0.5, 0.8 ± 0.5, 1.0 ± 0.6) or performing Methods 2-3 and calculating unbound density (mean % difference = 3.0 ± 2.2, 2.6 ± 1.8). However, mean coefficients of variation (CoV’s) in cone density calculated from simulated images using Methods 1-3 (37µm × 37µm ROI) were smaller for bound densities (0.7%, 0.8%, 0.7%) than for unbound densities (3.1%, 3.8%, 3.9%). Similarly, mean CoV’s in cone density calculated from in vivo cone images across eccentricities using Methods 1-3 were smaller for bound densities (2.4%, 3.1%, 3.2%) versus unbound densities (4.8%, 4.6%, 4.8%).
Computing bound cone density after marking all cones fully and partially within the ROI (Method 1) or fully within and partially along 2 adjacent borders of the ROI (Methods 2, 3) provide measurements with greatest accuracy and least variability. We suggest calculating bound density after performing Method 1 to provide maximum sampling, particularly for analyses done at peripheral eccentricities.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only