Keratoconus detection from videokeratoscopic height data

Purpose. Current videokeratoscopic-based keratoconus detection schemes primarily look for asymmetries in the dioptric power of the cornea. Recently, Hubbe and Foulks (Ophthalmology. 1994;101:1745-1748) have shown that poor fixation of normal patients can produce similar dioptric power asymmetries resulting in a false detection of keratoconus. A technique to detect keratoconus in the presence of corneal misalignment is investigated. Methods. Corneal height data for 61 myopic and 17 keratoconus patients is obtained with a TMS-1 videokeratoscope (Computed Anatomy, New York, NY). A 6 mm diameter zone from each set of height data is expanded into the orthogonal set of Zernike polynomials. Expansion terms representing a planar tilt fit the misalignment of the cornea, while the higher-order coma-like terms fit the cone. The magnitude of the first coma-like Zernike term and the spherical equivalent power obtained from the expansion coefficients are compared between the myopic and keratoconus groups. Results. The net "coma" coefficient is 0.0076 ± 0.0046 for the keratoconus group and is 0.0013 ± 0.0008 for the myopic group. These values differ significantly (p<0.0001) between the two groups, however the coefficients of four keratoconus (23.5%) lie within two standard deviations of the myopic mean value. A significant difference (p=0.0008) in spherical equivalent power between the two groups also results. Conclusions. A keratoconus detection scheme that utilizes corneal height data and is insensitive to poor patient fixation during videokeratographic screening may be useful in reducing misdiagnosis. The purposed technique is currently limited in its accuracy. While the net coma and spherical equivalent power between normal and keratoconus corneal height data differ significantly, a more thorough analysis of higher-order expansion terms and their interactions when fitting the cone is necessary to improve the usefulness and accuracy of this detection scheme.