Nucleus Segmentation in Automated Cell Microarray Image Analysis

Live cell microarray technology [1], allows the simultaneous analysis of many gene products. These microarrays are growing cells in spots printed on a glass slide using a robotic arrayer. The cells growing on the DNA and gelatin spots express the DNA and divide 2–3 times in the process of creating a microarray with features consisting of clusters of transfected cells. High-resolution images are obtained from fluorecence microscopy of different channels scanned from each spot. Nuclei, cytoplasm and membrane are expressed, depending on the purpose of the assay. There are many difficulties to automate this process that are common in clustered cell images: they form a complex structure with many touching cells, cells occur at different focal planes, and there is uneven background. Quality results in detecting the nuclei can only be achieved using sophisticated, but fast image analysis techniques. Speed is required because a microarray slide can consist of thousands of spot samples, each one with about eight 3-channel high-resolution images. The segmentation of the nucleus in the reference image is the main challenge of the image analysis. Our interest here is the performance the nucleus segmentation in the signal response estimation process. In this paper we investigate the role of the nucleus segmentation in the automated image analysis to estimate the signal response in nucleus-expressed assays. The procedure consists in segment the nucleus image marked by DAPI and use this segmentation to measure the gray-scale value in the nucleus-expressed image. We use two morphological segmentation techniques, one based on the top-hat plus separation of touching cells, and the other based on the gray-scale inner and outer marker watershed-based segmentation. We compare the effects of these two segmentation algorithms in the prediction of the signal response of a nucleus-stained assay. We conclude that the gray-scale watershed segmentation results in a more accurate segmentation.