Purpose: To evaluate the precision and repeatability of a Placido disc-based computerized videokeratoscope, using convex surfaces of varying eccentricities and apical radii designed to simulate the range of topographical variations of the human cornea, rather than the purely spherical surfaces used in most previous studies.
Methods: Form Talysurf analysis was used to verify the exact form of 12 Perspex convex surfaces. The EyeSys model II videokeratoscope was used to measure the sagittal radii of curvature twice at known points on each surface. The raw data tables were analyzed to assess the repeatability and accuracy for both central and peripheral points on each surface. The relationship between these factors and the eccentricity was investigated.
Results: For central radii the instrument showed high correlation (r = 0.996) between actual and measured values. There was a small instrumental bias of +0.042 mm and the 95% limits of agreement were narrow (+0.121 to -0.037 mm), indicating clinically acceptable accuracy. The accuracy decreased slightly as the p-value decreased (greater peripheral flattening). For peripheral radii, the overall accuracy compared well to central radii, with an average bias of +0.022 mm and maximum error in 95% of cases 0.083 mm (bias +1.96 x SD). However, for surfaces where p = 0.50, the bias was +0.049 and maximum error in 95% of cases 0.110 mm. Repeatability for the aspheric surfaces was shown to be high (SD +/- 0.01 mm in all quadrants).
Conclusions: The accuracy of the EyeSys corneal analysis system (CAS) in measuring central and peripheral radius of curvature was shown to be dependent on the shape of the surface to be measured. For more rapidly flattening surfaces, a decrease in accuracy was found for both central and peripheral radius of curvature, which in clinical terms is thought acceptable.