Keratoconus detection based on videokeratoscopic height data.

Abstract

To develop a videokeratoscopic-based keratoconus detection scheme that avoids the ambiguity of dioptric power definitions and videokeratoscope design. Corneal height data obtained with a commercial videokeratoscope are decomposed into the set of orthogonal Zernike polynomials. Expansion coefficients of a "normal" group and a keratoconus group are compared to find significant differences. Elevated Zernike terms are used to detect the disease in these populations. The performance of this detection scheme is compared to other videokeratoscopic keratoconus Indices. Two low-order Zernike polynomial terms are identified as being elevated in keratoconus patients and combined to form a new detection index. This index performed at least as well as keratoconus detection schemes based on the inferior-superior (I-S) value, the steepest radial axes (SRAX), and the Surface Asymmetry Index (SAI) for the samples studied. The proposed Zernike scheme offers a potentially viable algorithm for detecting keratoconus that avoids the ambiguities of dioptric power definitions and is independent of videokeratoscope design.