Research On Overlapped Protein Spots Segmentation For Wo-dimensional Gel Electrophoresis Images

Proteome is one of the important research contents of post genomics, and it’s ofgreat importance for exploring the mystery of life. It’s also meaningful in theidentification of diseases and the effect of drug in the process of life. Two-dimensionalgel electrophoresis is the key of analyzing the expression of proteome. With thecontinuous development of biological science and medicine, computer analysis of gelimage has become an indispensable technology. Computer-based analysis of gel imagesmainly includes image preprocessing (such as noise removal, background correction),detection of the protein, and gel matching, etc. The detection of protein spots is one of themost important steps in the analysis process. The quality of detecting and separatingprotein points from the gel electrophoresis image by using effective segmentationalgorithm is of important significance for the following analysis of protein spots. And itcan lead to accurate estimation of spots properties, e.g. spot volume. The followingprocedure (quantitative, matching and analyzing of the protein) depends highly on theprocess of detection. This thesis mainly studied the separation of the overlapped gelelectrophoresis image.The existing methods of protein spots detection mainly include watershed-basedmethod and shape-feature-based method. Based on the spots pretreatment on protein, thisthesis proposed a segmentation method for overlapped proteins. It’s based on thealgorithm of region tracking and minimum perimeter polygon (MPP) and it’s useful inthe segmentation of overlapping protein spots.The main work and achievements of our thesis is as follows：(1) The algorithm of pre-processing of the2DE is studied. The gel electrophoresisimage was pre-processed using spatial filters such as median filter, mean filter and gauss filter. The improved Non-Local means algorithm was also used. The improved Non-Localmeans can achieve gel image denoising based on the degree of similarity between pixels,which can protect the details and edge of gel image while denoising.(2) The algorithm of pre-segmenting of the2DE is discussed. Edge detectionoperators and the boundary tracking algorithm in the process of image pre-segmentationis used respectively to get the initial contour. The algorithm of boundary tracking startsfrom a point on the edge, and then connects all the edge points by searching and labelingall the edge points. It can obtain a relatively accurate edge, and can condense the edgeinformation. Meanwhile this algorithm has some denoising effect. The closed edgecurves obtained hereby can be more effective in the statistical analysis of the shape andsize of the protein points.(3) Polygonal approximation method is proposed in segmentation of overlappedprotein spots. Determination parameter is used to extract the overlapped protein spotsbased on the area and perimeter. The minimum perimeter polygon approximationprinciple is used to approximate the edge of overlapped protein spots. Then the concavityand the inner angle are determined based on linear behavior of the polygonal. Theconcave points obtained are matched with the matching principle, and the connectinglines are also confirmed. The overlapped spot points are then separated effectively. Thealgorithm is based on the edge obtained with boundary tracking algorithm to obtain thedistribution of independent protein spots, which can lay the foundation for subsequent ofresearch protein point.