Accuracy and reproducibility of KeraScanner analysis in PERK corneal topography

Abstract
The measurement of the corneal radius of curvature centrally and in the corneal periphery is exceedingly difficult because of variables in photography and data acquisition. We present a technique of Automated image scanning of corneal photographs which provides a more accurate and reproducible analysis of the paracentral corneal contour. We analyzed the sources of error of corneal topography measurement. CorneaScope photographs of calibration balls were generated by each Prospective Evaluation of Radial Keratotomy (PERK) Surgical Center. The 90% confidence intervals for individual ring-radial positions on the 8.00 and 10.00 mm balls were 7.91 to 8.09 mm and 9.89 to 10.11 mm respectively. Our isopter method of analysis for clinical comparison averages all eight radii of curvature for each CorneaScope ring. This isopter technique reduces the corresponding 90% confidence intervals to 7.96 to 8.03 and 9.95 to 10.06 mm. The measurement variability of a 10 mm calibration ball decreased in the following order: calibration ball positioning, scanner personnel, photograph movement between scans, and nonrepositioned scanning. Photographs of calibration balls demonstrate .+-.4 microns of measurement resolution and patient corneal photographs may be evaluated to .+-.40 microns. This automated scanning system provides consistent data when digitizing keratographs.