Automatic texture segmentation using morphological filtering on images of the human cerebellum

In microscopic images of the human brain the cerebellum appears as several different layers of varying textures. These different tissue layers are defined by the different cell types and cell densities that are observed within them. In order to be able to quantify certain aspects of neurodegeneration it is desirable that one can automatically identify these different cerebellar layers. It is of particular interest to the authors to be able to quantify the amount of cerebellar atrophy and to be able to automatically recognise differing distributions of particular pathological features with respect to the different layers of the cerebellum. The use of quantitative image processing techniques in biomedical image analysis is an increasingly competitive field of research. The application of these approaches to neuropathology is now a reality, with a number of different research groups reporting quantitative techniques for analysing a variety of pathological features. As yet few researchers have described systems which can accurately relate these features back to the underlying neuroanatomy. Towards this goal the authors have developed a novel image processing system which can segment images of human cerebellar tissue using mathematical morphological functions