BACKGROUND The image-forming properties of a keratoconus eye are degraded even in the early stage of this disease. The purpose of this study was to develop keratoconus detection scheme based on topography height data independent of the currently used system which avoids the disadvantages of detection algorithms currently used in clinical practice. PATIENTS AND METHODS Eighty-eight patients with keratoconus (46 with mild and 42 with severe clinical signs) and a control group of 40 normal subjects were included in this study. A decomposition of corneal topography height data into orthogonal Zernike polynomials was performed using the commercially available corneal topographer TMS-1. Expansion coefficients of the different groups were compared to evaluate significant differences. Elevated terms were used to detect the disease. The statistical significance of this detection scheme was compared to those given by the keratoconus detection software of the TMS-1. From the elevated Zernike terms a neural network was constructed and optimized for dividing keratoconus patients and normal controls. RESULTS Some low-order Zernike coefficients with a radial order n < 8 were found to be elevated in patients with keratoconus and were used to define a new detection algorithm. This index performed at least as well (sensitivity in mild/severe keratoconus 93.4%/100% with a specificity of 100%) as keratoconus detection schemes based on the Klyce-Maeda and the Rabinowitz-Klyce indices as well as the I-S value and the Surface Asymmetry Index SAI in our study population. CONCLUSIONS Zernike decomposition of corneal topography height data allows a definition of an exact and robust algorithm for detection of keratoconus. It avoids the drawbacks of refractive power based definitions and is independent on the individual topographer design.