Research On Cell FISH Image Segmentation,Counting And Retrieval

As far as digital image processing and cell genetics are concerned,that they are combined,would solve the key problem of analyzing cell morphological information on cell level,which is especially helpful in determining the quality of the mammalian germ cell.This research is significant to such applications of cell genetics as livestock breed,diagnosis of disease,chromosomal lesion,and so on.Because the resolution of the cell image observed by traditional microscope is low,and its sizes and shapes are various,and they usually cover each other,it is difficult to process cell image.Nowadays,there is no general way to segment,count and retrieve cell images.Therefore,it is necessary to find a fast and accurate method for cell segmentation,counting and retrieval.Based on pulse coupled neural networks(PCNN)model and particle swarm optimization(PSO)algorithm,this thesis commits to segment,count and retrieve cells fluorescence in situ hybridization(FISH)images.The main contributions are as follows.(1)The overview of fluorescence in situ hybridization in agricultureThe histories and technological details of FISH are summarized,and its application in the field of agriculture especially the animal husbandry is introduced,which lays the foundation of the following cytogenetic technology sections(2)The mammalian germ FISH image segmentation and cell counting based on PCNNThe image segmentation of integrating watershed algorithm with PCNN,and the cell labeled method are proposed,which can count cells FISH accurately.Firstly,cell FISH image is denoised by PCNN and local median filter;secondly,rough image segmentation is implemented by watershed algorithm,and adhered cells are separated;thirdly,explicit segmentation is implemented by PCNN with the minimum cross entropy;finally,the segmented cell images are labeled,and their numbers are counted.Experimental results show that the accuracy rate of cell counting reaches at 93.3%.(3)FISH image retrieval based on improved PSO algorithmTo overcome the shortcomings,large calculation and low precision,of traditional content-based image retrieval algorithms,the improved PSO algorithm is put forward for cell FISH image retrieval.Firstly,the template image and images in database are blocked,and their histogram information is extracted;secondly,the template image retrieval is equal to classification by PSO;finally,the results of image retrieval are those with the most similar images.The experimental results show that the proposed algorithm,based on PSO and the novel classification algorithm,could quickly find out a similar image groups from the public image database,and the average retrieval precision and recall rates reach 88% and 93.3%,respectively.The average time consuming is only 3.48 s,and applicable in cell FISH images in image retrieval.There are two kinds of signification about this thesis.On the one hand,it can meet the development requirements of biology,and make it automatic to release the burden of biologists;on the other hand,this thesis is of interdisciplinary exploration,and digital image processing technics are extended to cell FISH image segmentation,counting and retrieval.