The use of gray-level information and fitting techniques for precise measurement of corneal curvature and thickness

This paper describes several methods to compute corneal curvature and thickness from slit-lamp images. The methodology to recover corneal shape from noisy data is based on original contour point extraction algorithms using gray levels and a threefold fit to the extracted points including a least-squares fit, low-pass filtering of the residue, and splines adjustment to this filtered residue. The results of both simulations and image processing of corneas in living eyes show that those methods improve the repeatability of the measurements as compared with usual processing techniques. Final results (40 μm for curvature repeatability and 30 μm for thickness) are discussed and limiting factors are analyzed.