Purchase this article with an account.
Suman Adhikari, Hope M Queener, Gwen Musial, Hanieh Mirhajianmoghadam, Jason Porter; Evaluating different region of interest sizes for quantifying cone density using simulated and in vivo retinal images. Invest. Ophthalmol. Vis. Sci. 2020;61(7):475.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Different retinal areas, or regions of interest (ROIs), have been used to calculate cone metrics in vivo, making it challenging to know whether it is possible to compare metrics derived differently across studies. We sought to better understand the impact of ROI size on the quantification of cone density.
5 square ROIs of random orientation for each of 5 different sizes (25, 37, 50, 75, 100µm on a side) were extracted from simulated cone mosaics with gradient density profiles based on Curcio et al.’s (1990) data along the temporal meridian at eccentricities of 0.075 - 2mm. 5 ROIs of random orientation for each of the same 5 sizes were also placed on in vivo cone mosaics from 5 healthy adult subjects acquired using an adaptive optics scanning laser ophthalmoscope at eccentricities of 0.3 - 2.1mm along the temporal meridian. A custom program (Mosaic Analytics) was used to calculate bound cone densities for all ROIs. A two-way ANOVA was used to compare density measurements between ROI sizes.
For simulated data, the 25µm ROI size yielded values that were significantly different from all other ROI sizes and were closest to simulated values at 0.075 and 0.1mm eccentricities of (mean % error ± SD = 0.05 ± 0.30%, P<0.01 and 0.27 ± 0.39%, P<0.05, respectively). However, there were no significant differences in densities measured by different ROI sizes at all other eccentricities (0.2-2mm). Coefficients of variation (CoVs) in calculating cone density increased with increasing eccentricity for 37, 50, and 75µm ROI sizes. Across eccentricities, CoVs were most variable for small ROIs (mean = 0.9 ± 0.4% and 0.5 ± 0.4% for 25 and 37µm) compared to larger ROIs (0.2 ± 0.1% and 0.2 ± 0.1% for 75 and 100µm) (P<0.01). Consistent with simulated data, there were no significant differences in densities measured by different ROI sizes from in vivo data at all examined eccentricities (0.3-2.1mm). The CoV for the 25µm ROI across all eccentricities (mean = 14.8 ± 6.3%) was more variable than all other ROI sizes and increased with increasing eccentricity.
For small eccentricities, computing bound densities using the smallest ROI size provides measurements with greatest accuracy. While our data suggest that bound density calculations for eccentricities >0.2mm are independent of ROI size, we suggest using larger ROI sizes due to their lower variability in healthy eyes.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only