Research And Application Of Comprehensive Method For Intelligent Analysis Of Chromosome Image

Chromosome image processing,is the key step in the medical imaging and human genetics research,and is an important method to assist doctors in clinical diagnosis,but also a prerequisite to detect various genetic lesions from the abnormal chromosome morphology(such as translocation,repetition,missing or different length,etc.).However,it's too complex for requiring manual operation of the computer,which means a low degree of intelligence,and the effect may be disturbed by human factors.Therefore,from the perspective of the application of chromosome image processing technology,this thesis introduces,implements,combines and improves the algorithms that need to be used.Mainly involved in computer graphics,the depth study of image segmentation and classification technique.Part of graphics includes bone extraction,bone pruning,image enhancement and image deformation algorithm,which provides the preprocessing means for segmentation and classification network.Part of image segmentation implements the transformation from the initial images to a single or clusters of chromosome conversion,discards the noise information,and extracts the key parts of the whole image.Part of image classification is used to recognize straightened chromosomes by deep network.In addition,this thesis will also make use of the above algorithms to realize a prototype system.This thesis is focus on the graphics algorithm.This part migrates and optimizes related algorithms from other application fields,adding additional steps and intelligent parameters to reduce the need for manual intervention.The innovation point is to try to use the result of deformation processing,to assist the existed deep learning model,so as to enhance the accuracy of the network.