Research And Implementation Of Statistical Shape Model Using Normal Liver Cell Images

Pathological diagnosis is called the golden standard of clinical diagnosis.But the number of pathologist face a serious shortage,at the same time,heavy wordload is also a problem.So,there is an urgent demand for the computer-aided diagnosis(CAD)technology based on pathological images.However,the CAD technology based on pathological images lags far behind CAD technology based on CT and MRI.The liver cell images segmentation is a primary problem of patholossgical liver tissue diagnosis,and for cancerous liver cell recognition and later diagnosis,building a statistic shape model using normal liver cell images has a significance of research and aplication.After studying a large number of relevant literatures and an indepth research,consider the liver tissue pathology image as research object,studied the liver cell segmentation method and statistical shape model based on the normal liver cell images.Finally,a novel segmentation algorithm for overlapping cells and statistical shape model were proposed in this paper.Specific as follows:(1)Liver cell segmentation is the prerequisite of building liver cells statistical shape model.There are three parts in this paper about the segmentation algorithm including image preprocessing,coarse segmentation and overlapping cells segmentation.Mean filtering and histogram enhancement based on R-channel are used in the image preprocessing stage.In the process of coarse segmentation,level set algorithm is the main method.An improved watershed algorithm based on adaptive neighborhood was proposed in this paper to segment the overlapping cells.The algorithm can not only correct the result of conventional algorithm but also improve the efficiency of conventional algorithm.(2)In the part of building normal liver cell statistical shape model,first,the liver cell shape is showed in polar coordinates.Then the cell shape images alignment problem can be solved by transferring from two-dimensional to one-dimensional.As a result,it is not only able to simplify the process of landmark point acquisition but also improve the efficiency of statistical shape model.To verify the effectiveness of the algorithm,we perform the experiments of cell segmentation and establishing statistical shape model building for HE straining liver tissue pathology image.The experiment results show that the proposed algorithm can get the better segmentation results and establish a good cell shape model.Based on the above algorithm,the system of segmentation and statistical shape model is constructed in this paper.The system can not only provide a good user interface but also provide an experimental platform for further research.