Abstract
Purpose: :
To examine the relationship between conjunctival redness and roughness, each measured objectively, after determining the repeatability of the metrics to be used.
Methods: :
Digital images of the tarsal conjunctiva of 29 subjects (each image with sufficient redness and specular ["roughness"] information) were quantified after separating the redness and roughness components. Fractal dimension and mean and median RGB red intensity, and 13 metrics of roughness (Chinga et al., 2007, J Microscopy, 227(3) p254), were used to characterize the red and specular components respectively and CIE u’ was measured using a SpectraScan650 spectroradiometer. Each image was processed (using ImageJ) twice by a technician naive about the experiment’s purpose. Repeatability statistics were derived for each metric and the 3 most repeatable roughness metrics were compared to the redness metrics to determine the strength of the association between redness and roughness. Data analysis was done using R.
Results: :
The redness metrics were repeatable, with intraclass correlation coefficient (ICC), correlation coefficient of concordance (CCC) and Robinson’s A ranging from 0.735 to 0.983. The range of the same repeatability statistics for the roughness metrics was 0.496-0.995. There was significant correlation between redness and and 2 of the 3 repeatable roughness estimates( all r≥0.47, all p<0.011). Canonical correlation between the redness and roughness metrics was 0.885.
Conclusions: :
There were a number of metrics that provided repeatable estimates of conjunctival roughness from the specular reflections in tarsal images. The most repeatable was surface profile area. In addition, there were significant associations between redness and roughness metrics. Perhaps, only one aspect (redness or roughness) needs to be be quantified when objectively assessing the tarsal conjunctiva.
Keywords: conjunctiva • cornea: tears/tear film/dry eye • imaging/image analysis: clinical