Research In Segmentation Algorithm Of Cells Two-photon Microscopic Image

Using microscopy technology to analysis human cells is an important method for pathological study, diagnostic etc. As a pioneering imaging technology, the two-photon microscopy has many advantages:reduced photo toxicity and photo bleaching as well as enhanced imaging penetration depth. Therefore, the two-photon microscopy image is far more detailed than ordinary microscopy image. In the analysis of the cell image, image segmentation is one of the most crucial procedures. There are many segmentation methods for cells microscopic image, but few of them were aimed at cells two-photon microscopic image, the segmentation results of these methods cannot meet the requirements of precision analysis in cells two-photon microscopic image.This dissertation focused on the effective segmentation of cells two-photon microscopic image. The proposed processing method including segmentation preprocessing, cells center location and level set method for extracting the edge of nucleus. Among which the key step is the cells location. According to the priori knowledge of cell shape and the principle of template matching, this dissertation proposed two location methods by searching for round or oval shape-much like humans visual system do, one is based on roundness and the other is based on random sample consensus (RANSAC). Finally, based on the location result, reformed distance regularized level set evolution (DRLSE) was used to extract the precise edge of nucleus. The experiment results revealed that the proposed location method is efficient and the edge of nucleus is accurate, which can provide a useful basis for the automatic quantitative analysis of cells two-photon microscopic image.