Research On Faint And Overlapped Protein Spots Detection In2D Gel Electrophoresis Images

The extraction of differential protein spots is one core content in the2D-GE imageanalysis, while protein detection is the key step for extraction of differential proteinspots; provide the basis for disease diagnosis, drug research and analysis ofenvironmental pollution. Currently, how to correctly identify faint and overlappedprotein spots on the gel image is a hot and difficult point in the field of2D-GE imageanalysis. The detection of faint and overlapped protein spots is studied in depth in thispaper, which had been supported by the project of National Natural Science Foundationof China, Natural Science Foundation of Jiangxi Province and the graduate studentInnovation Foundation of Jiangxi Province. The mainly working content and researchachievements are as follows:The preprocessing method of2D-GE images was studied. In order to eliminatenoise interference on the gel image, this paper studied NLM method and compared toother methods such as traditional space, morphological and wavelet, and throughexperimental verification NLM denoising method has better denoising effect. Toeliminate background uneven of gel image, top hat transform method was used tocorrect image background, by the uneven illumination and interference stripe imageexperiment, verified the validity of this method.An initial protein spots detection algorithm based on image enhancementwatershed was presented. Firstly, In order to effectively detect the faint protein spots,this paper used top hat transform and CLAHE enhancement algorithm on the gel imageto improve the contrast of protein spots area. After that, the marker control watershedmethod was used to detect the initial outline of the protein spots. The results of theexperiments on different types of real gel images show that this algorithm has highercorrect detection rate, and effectively improve the faint protein spots undetectedproblems, but there is still difficult to separate the overlapped protein spots, furtherresearch is needed.A segmentation algorithm based on shape characteristics for overlapped proteinspots was proposed. Firstly, the initial detection algorithm based on image enhancementwatershed was used to detect the initial outline of the protein spots. Secondly, shapemarkers were adaptively extracted by gray-scale information and shape characteristics. Then, the exact Euclidean distance transform algorithm was adopted to calculate theshape distance image according to shape markers. Finally, the watershed algorithm wasapplied again on the shape distance image to separate the overlapped spots. The resultsof the experiments on different types of real gel images show that the algorithmeffectively improves the problems of faint spots undetected and overlapped spotsunseparated, and has higher separation accuracy, which is better than PDQuest8.0.1andaccumulated gradient watershed method, can meet the requirements of practicalapplication.