Abstract
Purpose :
Corneal opacification is a leading cause of preventable blindness globally. The emergence of novel therapies urges the development of objective quantitative measures to assess corneal disease. We aimed to develop an image analysis technique quantifying corneal opacification to measure the effect of a novel treatment in a mouse model of microbial keratitis.
Methods :
C57 mice had one eye inoculated with Pseudomonas aeruginosa and were allocated into 1 of 3 treatment arms: standard treatment, vehicle control and vehicle+treatment. Over 14 days en-face 24-bit colour photographs of the cornea were captured with a SPOT RTKE camera (Diagnostic Instruments) connected to a Leica MZF III stereo Microscope. Two ophthalmic clinicians acting as masked independent observers analysed 326 images, in 4 batches in the same randomised order. The area of opacification was measured using Fiji, an open source image-processing package based on imageJ. Definitions of corneal opacification, adequate and inadequate images were agreed upon prior to commencement of image analysis by the observers.
Measurements, in mm2 were plotted using ggplot2 in R, and loess smoothers fit to the chronological series for each assessor. The randomised order dictated that there should be no time-trend in the measured areas. We expected the average area measured to remain constant throughout the experiment for each assessor, and the assessors to show reasonable agreement.
Results :
Twelve images were deemed inadequate by observer 1 (in red) and were excluded from analysis.
Analysing absolute values (Figure 1), initially the observers were in reasonable agreement but this decreased with time. The vehicle (group 4) and vehicle+treatment (group 6) demonstrated less opacification, in mm2, compared to the conventional treatment (group 2) (Figure 2a & 2b) by the end of the study. While the actual values differed, but both observers demonstrated similar trends.
Conclusions :
We have developed a simple protocol to measure area of corneal opacification in our mouse model. Time dependant fatigue, image quality and image analysis training contribute to the variability of the measured corneal opacity. The initial agreement suggests that harmony can be achieved and the consistent trends are noted in each of the treatment groups demonstrating the value of this tool.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.