Contour detection of the cornea from OCT radial images

In this paper, we present a new approach for contour detection in the frame of OCT images of the cornea. With the aim of detecting upper and lower curves that define the cornea in radial slices, we have elaborated a three step specific method. The first step consists in finding epithelium and endothelium points (we call them anchor points). Then the image is preprocessed to enhance the contrast and to reduce speckle. In the final step both anchor points are used as starting points for the algorithm: we trace the contour of the cornea pixel by pixel from these two points with a weight criterion. Taking into account geometrical knowledge of the cornea enhances the robustness of the technique. We validate our approach on a bank of scans including low quality images.