Research On Overlapping Protein Spots Detection For Two-dimensional Gel Electrophoresis Images

Proteomics has been increasingly utilized in a wide variety of biological problems.The Two-Dimensional Gel Electrophoresis(2-DE)technique is a powerful proteomics technique.The goal of 2-DE image analysis is to rapidly identify proteins located on a single gel and differentially expressed proteins from a series of 2-DE image samples.Protein spot detection is the fundamental components of 2-DE image analysis.Due to the complexity of 2-DE images and the presence of artifacts,the detection of protein spots in these images is non-trivial,and time consuming.So the protein spot detection remains arduous and difficult.To separate the overlapping spots of protein in 2-DE images,this thesis proposed two segmentation methods for overlapped proteins.The main work and achievements of our thesis are as follows:(1)The pre-detection algorithm of the 2DE was studied.A comparative study about spatial filters such as median filter,mean filter and gauss filter was carried out,thus the Gaussian filter algorithm was used for gel image de-noising to reduce the noise effect on the subsequent protein gel image detection;A comparative study about protein point pre-detecting based on threshold value algorithm,watershed algorithm and region tracking algorithm was carried out.The region tracking algorithm was thus adopted for gel image protein point pre-detecting and clear boundary extracting of protein point,which laid the groundwork for subsequent overlapping point of protein separation.(2)The algorithm of overlapping protein segmentation based on the matching concave region was proposed.Firstly,the convex closure structure of overlapping spots was acquired in the gel images,and concave areas of overlapping spots was gained by subtraction of convex closure and original binary image;Secondly,the concave regions were matched according to the overlapping models;Then theconcavity points were determined according to the shortest Euclidean distance between the matching concave areas;Finally,overlapping protein spots were separated by constructing the segmentation lines between pairs of concave points.The extraction of concave points on original boundary of protein point boundary is converted into the extraction of concave points on concave areas contour,so that the range of concave points searching,the number of pixels and the speed of segmentation were significantly reduced.Experiments show that the proposed separating algorithm can achieve better results where concave regions are obvious and not complicated.(3)The algorithm of overlapping protein segmentation based on the combination of corner detection and polygonal approximation was proposed.First,Harris corner detector was used to distinguish the corner points from the overlapping protein spots boundary;second,to reduce interference on protein point border pits,the corner points was regarded as the feature points.The corner points were used to make polygonal approximation of overlapping protein spots boundary based on the polygon approximation algorithm;then the concave corner points are obtained by judging the convexity and concavity of polygonal vertices(corner points).Finally pits matching criterion was put forward and the real concave points were selected from the concave corner points.The overlapping protein spots were separated accurately by constructing the segmentation lines.The algorithm is simple and does not need multiple corrosion and expansion operations,while maintaining protein edge point to the greatest extent.Experimental results show that the error of concave points extracted by this algorithm is less than that by the algorithm based on concave region matching.The proposed algorithm has high correct separation rate and significant effect on the situation of complex and severe overlapping spots.