Author(s)

K. Mele & A. Leonardis

Abstract

Detection of ductus in mammary gland tissue by computer vision K. Mele & A. Leonardis Computer Vision Laboratory Faculty of Computer and Information Science University of Ljubljana TriaSka 25, 1000 Ljubljana, Slovenia Abstract In cell nucleus images of mammary gland tissue, both features of the nucleus and nuclei group shapes serve as cancer diagnosis criteria. Also, further treatment is dependent on the information about the cell nucleus spatial arrangement formed by malignant, potentially malignant, and normal ductus. In this paper we present an automatic image analysis method that we developed for detecting the structures of nucleus clumps that represent boundaries of ductus. The method consist of the following stages applied in succession: segmentation with a threshold, greedy algorithm, relaxation, and graph search. We tested our algorithm on images of mammary glands tissue. The results indicate that the method can distinguish between healthy ducts that have regular shapes and the malig

Keywords