Research Of Pathological Image Process Based On Data Mining

At present,most of the pathology is based on the description of specific morphology.Because of the lack of quantitative criteria,the influence of subjective factors on the diagnosis results is very large,and the pathological diagnosis results are different.Based on the methods and theories of computer technology and medical pathology,the research and analysis of biological cell image is being paid more and more attention.The use of effective methods for automated or semi automated analysis of cell pathology image,feature extraction and classification and recognition,can help pathologists,doctors give accurate and rapid diagnostic results.In this paper,I use the technology of data mining to study the pathological images,and find the quantitative association rules.The main work of this paper is as follows:Construction of Cytopathic Image Feature Database.Based on the advantages of super-pixel segmentation and threshold segmentation,the paper constructs a cascade segmentation algorithm,which can be used to segment the pathological images effectively.Then,based on the visual difference between nucleus and impurity and the main characteristics of nucleus,the characteristics of optical density,morphology and texture were extracted,and the database of cytopathological images was obtained.Selection of Pathological Image Features.In this paper,the principal component of different dimensions of the cytopathological data is selected by the t-sne method.The dimension of the feature selection is determined based on the visualization results of the characteristic data distribution of the nuclei,impurities and dimensionality reduction.Based on the chi-square test,the influence factors of the discretization feature data on the category are calculated.It is found that the algorithm can get better results in efficiency and result with features of the top 20 influencing fators.Research on the algorithm of pathological image mining.The paper uses the coverage rate and accuracy for parameter optimization,and the optimal solution is obtained by using particle swarm optimization algorithm.Based on the binary particle swarm optimization algorithm,the association mining algorithm has been improved,which can get fast,low support,high reliability,long frequent items.Based on the Apriori mining algorithm and the optimization algorithm,the quantitative association rules between the cells and the impurities are obtained.