Classifying human endothelial cells based on individual granulometric size distributions

摘要

This paper presents an application to a medical problem of methods of shape analysis based on mathematical morphology. The medical problem consists on the detection of abnormalities in the corneal endothelium, a tissue composed by quasi-planar cells of ideally regular hexagonal shape. Images of this tissue are taken by a specular microscope and used to evaluate the corneal endothelium status. Up to now, cell density, hexagonality and an analysis of cell areas are the usual descriptors of a corneal endothelium. These parameters are not sensitive enough to detect subtle lesions. What this paper proposes is an analysis based on granulometries, which are size-shape descriptors widely used in Mathematical Morphology. Applications of granulometries lead to distribution functions whose moments are used as shape descriptors. Former approaches proposed a global evaluation of the whole corneal endothelium but irregularities affecting to a few cells could be ignored. Differently to that approach, this paper uses the granulometric size distribution of each cell. A group of normal cells is chosen as controls and a given cell is classified as normal or abnormal by comparing its granulometric size distribution with the corresponding distributions of the controls. The technique is illustrated with the analysis of some images of corneal endothelia.