Segmentation of confocal microscope images of cell nuclei in thick tissue sections

摘要: Summary Segmentation of intact cell nuclei from three-dimensional (3D) images thick tissue sections is an important basic capability necessary for many biological research studies. However, segmentation often difficult because the tight clustering in specimen types. We present a 3D approach that combines recognition capabilities human visual system with efficiency automatic image analysis algorithms. The first uses algorithms to separate into regions fluorescence-stained and unstained background. This includes novel step, based on Hough transform focusing algorithm estimate size nuclei. Then, using interactive display, each nuclear region shown analyst, who classifies it as either individual nucleus, cluster multiple nuclei, partial nucleus or debris. Next, morphological reconstruction watershed divides clusters smaller objects, which are reclassified by analyst. Once no more remain, analyst indicates should be joined form complete was assessed calculating fraction correctly segmented variety types: Caenorhabditis elegans embryos (839 correct out total 848), normal skin (343/362), benign breast (492/525), cancer line grown xenograft mice (425/479) invasive carcinoma (260/335). Furthermore, due analyst's involvement process, always known population not, assuming judgement correct.