Splitting touching-cell clusters on histopathological images
Hui Kong, Metin N. Gürcan, Kamel Belkacem-Boussaid · 2011
In this paper, we propose a novel algorithm for splitting touching/overlappingcells in histopathological images. Given a binary segmentation map by which the cell nuclei have been delineated and separated from the other regions, for each connected component, we differentiate whether it is a touching-cell clump or a single non-touching cell after we smooth out its boundary by Fourier shape descriptor. The differentiation is mainly based on the distance between the most likely radial-symmetry center and the geometrical center of the connected component. Finally a new iterative splitting algorithm is only applied to the touching-cell clumps based on detected concave point and radial-symmetric center. We tested our splitting framework on 21 challenging Follicle Lymphoma images and get an average error rate of 5.2%.