Research On Pathology Image Processing Algorithm Of Microscopic Cells

In recent years,the cell DNA quantitative analysis technology by means of computer to conduct automatic identification and diagnosis on the microscopic pathology images gradually replaced the artificial naked eye diagnosis and it has become the mainstream of cytological examination.Cell DNA quantitative analysis technology usually through the computer automatically analyze the microscopic cytopathic image,split the independent nuclei,and calculate the DNA Index to provide a reliable basis for for tumor diagnosis and identification.Computer-assisted quantitative analysis of cellular DNA can largely improve diagnostic efficiency,reduce the workload of pathologists and the possible error diagnosis and missing diagnosis.The results of DNA quantitative analysis of cervical cells were affected by many related factors such as cervical sample preparation,sample digital image acquisition,cervical cell nuclear segmentation,classification,DNA index algorithm and so on.In this thesis,we introduce the relevant influencing factors at the same time and conduct the in-depth study on the following content:Firstly,the accuracy of the different methods for calculating the nuclear optical density of cervical cells in the DNA Index algorithm was compared according to the principle of cervical cytology,and the mean value method,the median value method and the improved peak value method for the reference diploid cell selection in the DNA Index were studied according to the CV value standard.Then combine different nuclear optical density integral calculation methods of cervical cell and the diploid cell selection methods and use variance,repeatability error and other factors to evaluate the stability and accuracy of different DNA Index algorithm.After getting the DNA Index,correlation analysis was conducted to research the relationship between the distribution of DNA Index and positive and negative cervical samples in DNA quantitative analysis,and the DNA Index distribution event with high correlation with the positive samples was found.Then,such DNA Index distribution event was used as the classification criterion of cervical samples,and the effects of improved DNA Index algorithm on DNA quantitative analysis were evaluated by specificity,sensitivity,negative predictive value and positive predictive value.Finally,on the basis of careful study of the differences between cervical cancer nuclei and normal cervical nuclei,the related characteristic parameters were designed according to the morphological,grayscale and texture features of cervical lesion nucleus,the association rule mining method was used to find the characteristics of high correlation with cervical lesion nucleus,which was helpful for the classification of lesion nucleus in all of the cervical nucleus.