Researches On Endothelial Cell Segmentation And Design Of Segmentation Software

The endothelial cell is an important subject in a number of medical researchessuch as the pathogenesis and treatment of a variety of diseases, e.g., atherosclerosis.With the higher requirements on quantification of the experimental results, qualitativeanalysis of microscopic images of endothelial cells by human eyes is not satisfactory,and more objective and accurate quantitative analysis is desirable. In order to obtainthe required quantitative results with the method of the computer-aided analysis orcomputer automatic analysis, the research of endothelial cell microscopy imagesegmentation has the important theoretical significance and the direct value.Based on the different levels of the required manual operation in thesegmentation process of endothelial cell microscopy image, this paper gives threetypes of endothelial cells segmentation method: manual segmentation method,automatic segmentation method, semi-automatic segmentation method.The manual segmentation method is entirely dependent on the operation andcontrol of the segmentation process, and according to their visual judgment toevaluate the segmentation results to the satisfaction of the by-cell region of interestextraction. The manual segmentation method used alone for a region of interestthreshold segmentation method to achieve a single cell division, and through theinterface mode of operation designed to minimize the operation of the labor intensity.Semiautomatic segmentation method based on local tolerance region growing andtolerance region based on the global search method, by selected region growing seedpoint, four-connected region growing, according to the local tolerance set by the useras growth termination condition, according to regional intensity distribution of thesimilarity criteria to find to meet the global tolerance within the cell region.Automatic segmentation method is firstly filtered using morphological imagereconstruction filter, and pretreatment the low-contrast images of endothelial cellswith histogram equalization, and then use adaptive thresholding as a coarsesegmentation, combined with morphological dilation and erosion operation to buildthe markers of the cell region and background region, Finally used themarker-controlled watershed segmentation algorithm to complete the split andavoided the occurrence of over-segmentation.The above segmentation methods are carried out experiments on the actual imageand the experimental results show that the respective methods for certain types of images can effectively split.Based on the OpenCV image processing library, programed the abovesegmentation algorithm in Visual C++2005, designed and implemented theendothelial cell microscopy image segmentation software. Through the software, theuser can select the appropriate combination of manual segmentation andsemi-automatic segmentation and automatic segmentation method, in order to obtainthe required accuracy of segmentation results at the same time, reduce the laborintensity. The software not only can be used for the segmentation of microscopicimages of endothelial cells, and also for other types of microscopic imagesegmentation provides a good foundation and a number of applicable methods.