Study On Medical X-ray Image Segmentation Method Based On Improved Ncut

ABSTRACT:Medical image segmentation is the foundation of high-level medical image analysis and understanding. It has significant research value and application prospect in clinical treatment, medical study and some other research fields. Individual differences in human tissues and the limitations of medical imaging technology make medical image with blurred edges, various kinds of noise, artifacts and so on. This complexity of medical image makes its segmentation become a difficult issue in the field of image processing. Recent years, graph theory based image segmentation methods have been increasingly applied to medical image. Among all these graphic algorithms, Ncut criterion has attracted the attention of many researchers due to its ability to maximize the similarity in the same class and the dissimilarity in different classes simultaneously. What’s more, it has a rigorous mathematical method to solve the partitioning problem. After studying the research status in the field of medical image segmentation, this paper selectively analyzes graph based optimal cut criterions and their limitations. Then, the paper develops an in-depth study of medical X-ray image segmentation based on Ncut to segment masses in mammography and puts forward the implementation scheme of a medical image processing system. The main research work and innovate ideas are as follows:1. The paper studied the corresponding relation between characteristics in image and graph in graph theory based image segmentation method. Then, common graph optimal cut criterions had been introduced and compared in detail. According to the characteristics both in medical image and partitioning methods of graph, this article particularly analyzed the disadvantages of Ncut algorithm in its application and development in image segmentation field. This provides a direction for the subsequent segmentation algorithm improvement.2. Considering the peculiarities of Ncut including the sensibility to noise, intensive computation and single weight, this paper proposed a method named Shape-Controlled Ncut with Advanced Watershed (AW-SNcut) to make two improvements on Ncut algorithm:a) designed an improved marker-controlled watershed algorithm as pre-segmentation algorithm. It can mark interested region better and automatically select the size of morphological structure element. AW-SNcut regards each sub-region obtained from preprocessing as the input of Ncut algorithm in order to use regions instead of pixels to segment the image. Thus, the low computational efficiency problem of Ncut can be solved. b) Designed shape-controlled Ncut algorithm which introduces object’s shape feature into similarity matrix and solved the problem of single weight in traditional Ncut. The validation of the method had been verified in mass segmentation of mammography and simulation experiments had shown satisfactory results.3. After studying on Ncut based medical X-ray image segmentation, a mammography mass processing system had been designed. In order to satisfy clinical demands, the proposed paper put forward an overall architecture of the system and determined three functional modules:the basic function module, the segmentation module and the manual control module. The medical image processing system had been designed by utilizing Microsoft visual studio2010platform and OpenCV. The mammography mass processing system fulfilled functions including processing and preservation of mammography, creating electronic medical records, manual correction, etc.The proposed AW-SNcut method is capable to segment interested objects in medical X-ray images and it’s more efficient in computing than traditional Ncut algorithm. In this article, there are17illustrations,3tables and50references.