An Iterative Approach to Nucleus Segmentation for High Content Imaging in Cancer Research

Abstract

We present an iterative technique for nucleus segmentation in high throughput RNA interference (RNAi) screening. This technique acts as a crucial processing step towards cell segmentation and feature extraction. Our data comes from three-channel RNAi cell images, with the nucleus information contained in a single DNA channel. Accurate segmentation of the nucleus information provides valuable prior information regarding cell counts and cell positioning, and is thus a valuable prior toward the overall goal of phenotype recognition and detection. Our iterative technique takes direct advantage of image gradient information to obtain accurate nucleus segmentations. It is particularly effective in separating nuclei that are very closely spaced, that thresholding cannot accurately segment