OCT 28, 2013
This study found that the selection of particular explanatory variables of Zernike expansion coefficients of corneal topography in discriminant models may contribute to improving the accuracy of keratoconus detection more than the discriminant model used. The results indicated that appropriate selection of explanatory variables gave similar results despite the use of different discriminant models.
The study’s authors compared four statistical regression methods under identical conditions using Zernike coefficients of corneal aberrations with the goal of improving the ability to detect keratoconus patterns by corneal topography in clinical practice.
They studied 51 eyes with keratoconus, 46 keratoconus suspect, 50 following LASIK procedures and 65 normal eyes. Four statistical discriminant analyses—linear discriminant analysis, k-nearest neighbor algorithm, Mahalanobis distance method, and neural network method—were performed using Zernike coefficients of corneal aberrations obtained by a Placido-based topographer. The detection scheme was constructed using a training set of data from one half of the randomly selected study participants, and performance was evaluated by a validation set in the other half.
They found that performance of the four models was different when fewer than 12 explanatory variables were included. However, performance with a greater number of explanatory variables and using the 2nd- to 4th-order Zernike terms for both 4-mm and 6-mm pupils did not differ significantly among models and converged around 78 percent.