Research On Segmentation And Analysis Techniques For Immunohistochemical Microscopic Image

With the rapid development of immunohistochemistry (IHC) technology, immunohistochemistry has become the important supplementary means for diagnosis and research of clinical pathological. Because computer-assisted image processing technology has some advantages such as accuracy, objectivity and fast processing speed, it has become the trend of immunohistochemical micrograph analysis technology. However, Many scholars'research results in immunohistochemical micrograph automatic segmentation are main limited in the segmentation of some organizational structure with obvious features. Moreover, the analysis of immunohistochemical images is mainly based on a single parameter and the specific image. These methods above have their limitation resulting in the final analysis is not comprehensive, objective and accurate. Therefore, in order to assist doctors with accurate observation and quantitative detection of immunohistochemical color reaction intensity, the study of immunohistochemical micrograph related segmentation and analysis has become the current and future research focus of domestic and foreign scholars.This paper makes ovarian cancer immunohistochemical image as the main object of study focusing on segmentation techniques of the positive cells region and analysis methods of immunohistochemical expression. We make some innovations in segmentation and analysis methods and the main study achievements are summarized as follows:(1) Do research on pathology knowledge associated with immunohistochemical micrograph. The research includes immunohistochemical staining methods, microscopic characteristics and classification of immunohistochemistry image, which is the basis for the following study of correct segmentation and analysis technical.(2) Do research on theory and methods associated with image preprocessing and color space conversion. I do a lot of comparative experiments in image enhancement and choice of color space, and finally select the more suitable enhancement method and color space for immunohistochemical micrograph segmentation.(3) Do research on the segmentation technique of immunohistochemical micrograph. Three methods are proposed and improved as follows: The first method is automatic segmentation for ovarian cancer immunohistochemical image based on chroma criterion. In the RGB color space, the chroma criterion and ISODATA cluster algorithm was used to extract the positive cells region. The second method is automatic segmentation for ovarian cancer Immunohistochemical image based on YUV color space. First, the image is converted to YUV color space and three components are extracted. Then, the OTSU segmentation method is used to extract the positive cells region. The third method is segmentation for overlap adhesion cells based on the improved watershed algorithm. Because the positive cells region after segmentation may exist some overlap and adhesion cells, the segmentation algorithm based on watershed was improved to solve the problem. The first method and second method are the segmentation methods to extract positive cells region. Experimentals show a good segmentation, and finally we compare the advantages and disadvantages of two methods, pointing that they can complement and vadicate each other. The third method is to segment the overlap and adhesion cells region. Experiments show that most overlap and adhesion cells can be divided and the algorithm is efficient.(4) Do research on the analysis technique of immunohistochemical micrograph. An analysis technique is proposed based on combination parameters considering the two factors of positive expression evaluation: the percentage of positive cells and the degree of positive staining. The percentage of positive cells is calculated by an average positive stained area percentage (APSAP) method, which is the ratio of positive staining area and the image total area. The degree of positive cells staining is calculated by the average optical density, that is, the ratio of integrated optical density and positive staining area. Therefore, positive level index (PLI) is derived, which is only need to calculate two parameters: the integrated optical density and the image total area. Experiments show that this method is more simple, objective and accurate.