Purpose
To quantify the relationship between specular focal distance and image distortion, in order to determine a predictive model of cell counting error based on amount of deviance from true endothelial cell shape and size.
Methods
High resolution specular image of a standard calibration lens was captured in focus to establish baseline pachymetry of 0μm. Subsequent images captured 30μm, 40μm, 50μm, 60μm, 70μm, 80μm, 90μm and 100μm from the baseline were analyzed for changes in area caused by optical distortion. Analysis was carried out in a single-blind trial with 10 sample counts performed for each specular image to determine mean areas.
Results
Given a true area of 10,000μm2 for the baseline distance of 0μm, a negative correlation was found to exist between focal distance (F) and mean area (A). For F=30μm, A=9666μm2; F=40μm, A=9444μm2; F=50μm, A=9180μm2; F=60μm, A=9055μm2; F=70μm, A=8869μm2; F=80μm, A=8767μm2; F=90μm, A=8694μm2; F=100μm, A=8467μm2. Using simple linear regression, the cell density error (E) can be extrapolated based on deviance from true area. For F=10μm, E=1.26%; F=20μm, E=2.90%; F=30μm, E=4.58%; F=40μm, E=6.53%; F=50μm, E=8.12%; F=60μm, E=9.99%; F=70μm, E=11.92%; F=80μm, E=13.91%; F=90μm, E=15.98%; F=100μm, E=18.13%.
Conclusions
Cell counting error can be considered negligible at 0-29μm, non-negligible at 30-59μm and problematic at 60-100μm of focal deviance.
Keywords: 481 cornea: endothelium •
550 imaging/image analysis: clinical •
549 image processing