Morphological analysis of mammary biopsy images

In this paper we present an automatic system to analyse cells' nucleus in a given biopsy of mammary tissue which is cancerous. Identification and characterization of cells' nucleus provides enough information to diagnose a high or low grade cancer. This is a high interest differentiation for the pathologist to apply the correct therapy. Performing these measures by human observation is a hard, imprecise and subjective task. The algorithm we present processes the image in order to identify cells' nucleus over the rest of the tissue. Images are enhanced and segmented using morphological transformations. An ultimate erosion is used in two steps to separate cells' nucleus in contact. It is based in a combination of symmetrical ultimate erosion with directional ultimate erosion.