The Study Of Segmentation Technology On Tumor Microscopic Cell Image

It is very important for assisting in the work and improving the diagnostic accuracy of doctors to use computer on microscopic cell image. To analyze cell fast and accurately, the cell morphology and other feature need to be extracted, to be measured and to be statistical. It is the base of medical microscopic image to be digital for analyzing the relevant parameters in cell image. However, due to the complexity of structure of the human body, tissues and organs as well as the irregular shape of the differences in different cell types, there will be a very big difference in cell structure, shape, extent and arrange. In the process of stomach epidermis tumor cell image segmentation, various complicated cells will be faced. Meanwhile, adhesion is very serious between nucleus or existed between nucleus and gland or between gland. So it is difficult to segment these cells. The accuracy of cell segmentation and consumption of computer will have a direct impact on the feature.In view of the limitations of traditional image segmentation techniques, this paper presents some new segmentation method.On microscopic view, a kind of cell tracking segmentation algorithm based on relaxation iterative was proposed to find out the cytoplasm, cell nucleus and background. There was no over-segmentation and the inadequate segmentation. What's more, the debris is little without too much adhesion. Then the chain table was used to track the cell nucleus and cytoplasm. Finally, line scan algorithm was used in the scanning of the pixels of cytoplasm parallel so that the cell nucleus in cytoplasm of the same cell was found out, which was innovational and based for the cell feature extraction.On macroscopic view, a kind of cell segmentation based on the maximization of mutual information in effective information was proposed to process stomach epidermis tumor cell image. To face the special gray, this paper used the cell and gland regions of smaller gray to be segmented, while the background regions with large gray was to be segmented simply. It is to avoid finding maximization of mutual information in every number of categories. As a result, it saved a lot of time and computer resources and improved the program's operating efficiency. For the optimal segmentation goal of maximization mutual information, this paper searched for optimal threshold to construct a new threshold segmentation based on maximization mutual information in effective information. As the same number of categories, this algorithm would search more thresholds with more cells and glands. Finally, the algorithm used threshold segmentation to process the tumor cells and glands fast and accurately.