Abstract
Purpose: :
To develop a computer-assisted imaging method to objectively quantify corneal opacity and haze.
Methods: :
Wild type and vimentin-deficient mice were subjected to corneal alkali injury to elicit corneal fibrosis by day 14. Fibrosis was pharmacologically inhibited by withaferin A (2 mg/kg/d) by i.p injection. Mouse corneas were photographed digitally with the aid of a dissecting microscope. Digital images of corneas were assessed by blinded human evaluators and corneal opacity was quantified using an opacity scale 0 (clear) to 4 (opaque) (n = 8 samples/group). The computer-based analysis of opacity was developed on color theory. Here, white light is composed of equal representation of red, green and blue and we proposed that corneal opacity that renders the cornea white in color represents a deviation from the equal mixing of red, green and blue. In colorimetry, colorfulness is a measure of how far a specific color is from neutral gray, independent of brightness. Visualizing each pixel in an image as a point in 3-D space characterized by a red, green, and blue axis, each color component corresponds to a number in the range from 0 to 1. The red, green, blue triplets are projected to a 2-D plane defined by red and green axes. On such a plane, shades of gray map to the (r=1/3, g=1/3) coordinates while solid reds, greens, and blues map to the three corners of, (1,0), (0,1), and (0,0) respectively, of a color triangle. A point cloud can be produced by repeating this map for each pixel within a region of interest (ROI). Point clouds were converted to a scalar measure of opacity by measuring the mean distance of these colorfulness points from their mean. R01s representing equal sized rectangular segments covering the brown colored iris from three non-overlapping regions of the cornea (avoiding the central pupil due to underlying lens) were used.
Results: :
The opacity values for injured vehicle- and WFA-treated mice when plotted as a function of their percentile distributions revealed that corneas of injured wild-type mice raking in top 10th percentile for clarity were significantly different from even the lowest 80th percentile of vimentin-deficient mice (P< 0.008), the latter being similar to wild-type mice treated with WFA. Surprisingly, vimentin-deficient mice treated with WFA recovered from mild haze that was observed in vehicle treated mice (p<0.0001). Statistical comparison of percentile distributions for human scored data and computer-assisted image analysis revealed the greater sensitivity of the latter.
Conclusions: :
Computer-assisted analysis of corneal opacity using the color information is very sensitive in detecting tissue transparency. We believe this novel method for analysis of corneal transparency could find use in research applications.
Keywords: imaging/image analysis: non-clinical • trauma • regeneration