Research On Bone Tissue CT Image Segmentation Method Based On Graph Cuts Algorithm

Segmentation of medical image was the basis for the analysis and diagnosis of human organs and tissues,which provided good technical support for extracting information around the diseased tissue,assisting doctors in disease analysis and performing three-dimensional reconstruction.With the development of medical imaging technology,the resolution of the medical image had been increasingly improved,and the details of the image had become more and more abundant.The increase in image complexity made it difficult to obtain satisfactory segmentation results with existing segmentation techniques.Graph Cuts algorithm,which could quickly and accurately extract the target of interest in the image,was a new idea for medical image segmentation in recent years.However,the traditional Graph Cuts method was deficient in automation and robustness,so,there were still some limitations in applying this method to the segmentation of complex and fuzzy medical images.In order to improve the efficiency of Graph Cuts algorithm in dealing with complex medical image segmentation,a feasible method by combining some characteristic information of medical image was to adopt the preprocessing algorithm to generate the marker points of foreground and background automatically,thereby greatly reducing the time of interaction,and ultimately improve the efficiency of Graph Cuts algorithm.This research was important in three ways.First,since bone tissue CT images were significantly different from their surrounding tissues in grayscale feature,an improved Graph Cuts algorithm whose marker points would be generated automatically by threshold segmentation was designed based on the interactive image segmentation framework.By introducing grayscale feature information of bone tissue and combining user interaction into the improved Graph Cuts algorithm,it could not only be effectively used to segment CT images of bone tissue,but also improve the segmentation efficiency of Graph Cuts algorithm.Second,an improved Graph Cuts algorithm whose markers were generated with the help of morphology was proposed after carefully analyzing the shortcomings of the existing Graph Cuts algorithm whose markers were generated automatically based on threshold algorithm in the experiment.The algorithm was tested by CT images provided by a hospital in Wuhan,and was compared with the segmentation results of the current mature commercial software.The result verified the effectiveness and feasibility of the improved algorithm.Finally,based on analysis of the interactive segmentation framework of Graph Cuts and the needs in the clinic,the overall framework of bone CT image processing system was proposed.Under the Windows platform,the development of the system was implemented in combination with the MFC framework and the VTK library.The system mainly performed functions such as reading and displaying of medical DICOM sequence images,preprocessing image,segmenting image,and reconstructing image.These functions were tested with corresponding CT images in the end.