Comparison and supervised learning of segmentation methods dedicated to specular microscope images of corneal endothelium.

The cornea is the front of the eye. Its inner cell layer, called the endothelium, is important because it is closely related to the light transparency of the cornea. An in vivo observation of this layer is performed by using specular microscopy to evaluate the health of the cells: a high spatial density will result in a good transparency. Thus, the main criterion required by ophthalmologists is the cell density of the cornea endothelium, mainly obtained by an image segmentation process. Different methods can perform the image segmentation of these cells, and the three most performing methods are studied here. The question for the ophthalmologists is how to choose the best algorithm and to obtain the best possible results with it. This paper presents a methodology to compare these algorithms together. Moreover, by the way of geometric dissimilarity criteria, the algorithms are tuned up, and the best parameter values are thus proposed to the expert ophthalmologists.

Robust 3-D reconstruction of surfaces from image focus by local cross-sectional multivariate statistical analyses: application to human ex vivo corneal endotheliums.

The considered problem of 3-D reconstruction consists in computationally and passively recovering both topography and texture of a scene surface observed by optical sectioning with a limited depth-of-field imaging system (typically a conventional optical microscope). Throughout a sequence of registered 2-D images, the concepts of shape-from-focus and extended-depth-of-field involve recovering both topography (depth map) and texture image of the surface by researching in-focus information, respectively. Toward that aim, traditional approaches generally follow a 2-D sectional way and thereby fail to deal with noisy and disturbed acquisitions, quite frequent in transmitted light observations and of interest in this paper. Such examples are the acquisitions of human ex vivo corneal endotheliums from the medical issue addressed in this paper, which are mainly damaged by cellular fragments in the sample immersion medium and by emphasized contrast reversals. To achieve with such noisy and disturbed acquisitions, a new focus analysis is introduced that originally adopts a 3-D strategy throughout the image sequence. This method exploits simultaneously all available cross-sectional cues that effectively strengthens the robustness. More precisely, it locally performs multivariate statistical analyses over cross-sectional spatial windows so as to find sectional in-focus positions. Comparisons to state-of-the-art methods on both synthetic data and real acquisitions from the deal-with medical issue demonstrate the efficiency and the robustness of the proposed approach.