Quantitative Analysis Of Immunofluorescence Based On Digital Image Processing Technology

Immunofluorescence images are widely used in clinical and scientific research,and although current imaging techniques have been able to observe specific proteins in cells at the molecular level,they are limited by the existing image processing methods,which can only perform simple digital information extraction using commercial image processing software and cannot deeply analyze fluorescence images quantitatively.Common biomedical image processing basically uses manual methods for digital information acquisition,and this method has problems such as long time consumption,poor reproducibility,and low accuracy.At the same time,the difficulty of analyzing immunofluorescence images is exacerbated by the emergence of problems such as blurring and ghosting due to the diffraction angle limitation of the mercury lamp excitation light;excessive background violence and low contrast of images;interference of impurities during the experimental process;and the existence of nonspecific binding of antigen and antibody.On the other hand,in recent years,computer technology has been developing rapidly,and new theories and technologies such as convolutional neuron networks,deep learning,and genetic algorithms are changing day by day.Therefore,how to combine new methods in computer field with immunofluorescence image processing,expand the diversity of analysis methods and improve the accuracy of image processing has become a research area of keen interest.Radiotherapy is one of the three major treatments for tumors,and usually,tumor cells undergo functional and morphological changes in the proteins,organelles and nuclei that involved in cellular activities after exposure to ionizing radiation.This thesis focuses on the immunofluorescence image analysis method of organelles(primary cilia),nuclei and protein localization in tumor cells after ionizing radiation,and develops the quantitative analysis of immunofluorescence based on digital image technology:(1)using color to separate different color channels,using the maximum interclass variance method to separate impurities,and comparing the contour extraction effect of different boundary operators.By personalizing and improving the characteristics of existing operators,a quantitative analysis method of primary cilia immunofluorescence images is proposed,which is more accurate and efficient than the traditional manual measurement method;(2)Image pre-processing by setting thresholds to initiate the acquisition of digital information starting from coordinate positions,the iterative global threshold segmentation method effectively mitigates the discrete noise that caused by image overexposure,solves the problem of excessive co-localization determination within the nucleus,and the immunofluorescence co-localization results obtained by using this logic operation are more realistic than traditional methods;(3)Automatic identification and counting of tumor cell nuclei is designed based on the YOLOv5 deep learning detection framework,which realizes the dynamic identification and counting function of tumor cell nuclei after immunofluorescence staining and provides a new method for live cell imaging detection and analysis.