A Biologically-inspired algorithm for the segmentation of cell nuclei in high resolution histological images

Immunohistochemistry (IHC) images are of high resolution and are stained for ER, PR, KI-67 and p53. Image processing can serve an important role in the diagnosis of disease from histopathological data due to its ability to process and analyze whole-slide digital images. Most of the traditional algorithms do not perform segmentation at a low visual level as the spatial relationship between pixels is not often entirely utilized. We developed an algorithm designed to mimic the visual system that utilizes a set of image features and identifies discontinuities within each feature domain. These features are further combined using a concept in neuroscience to generate an intermediate image that is more amenable to traditional tools for performing nuclear segmentation