Automated extraction of nuclear boundaries

Abstract

We present an automated method for the extraction of cell nuclei in images of kidney tissue sections. There are two stages in the segmentation strategy: (a) coarse segmentation by thresholding and morphological processing, and (b) accurate extraction of the nuclear boundary by region growing. The segmentation procedure is performed in a global manner and computationally less expensive than local approaches. Results obtained by applying the algorithm to several tissue images confirm its reliability.