Research On Image Segmentation And Non-rigid Registration Of Prostate DWI Sequences Images

With the development of information technology, medical processing and analysis which plays an important method in disease diagnosis and treatment. As the two key techniques of medical image processing and analysis, medical image segmentation and image registration are the foundation and prerequisite for image fusion, image reconstruction and image visualization. In the clinical diagnostics, prostate DWI is a sequence images which are taken under different magnetic field strengths during a certain interval. In the process of imaging, drift and deformation of images caused by the patient's breathing and involuntary activity will affect the correct diagnosis of doctors. Therefore the prostate DWI images need to registration. The main contents of this paper include:First, the background, significance and research status of the image segmentation and registration are full description. And then describe the basic principle of image segmentation and registration, including the concepts, classifications and methods.Second, region growing method is used for prostate DWI sequence images segmentation. This paper introduces the method of region growing and growing process. And the results of experimental segmentation are compared with the manual standard segmentation results, to evaluate the performance of the algorithmThird, In order to solve the problems of drift and deformation exist in the prostate DWI sequence images. A non-rigid registration method based on gray called Demons algorithm is applied in this paper. The Demons algorithm is described in this paper and applied to prostate DWI sequence images registration. Taking into account the measure of gray in energy function of Demons algorithm is single. When the gray value is small or zero, it is easy to cause mismatch. In this paper propose measure of curvature add to energy function. Curvature as a second order function contains more information. So that the image driven deformation under double measures of gray and curvature. The method has been applied to the prostate DWI sequence images registration. Through experimental comparison with original algorithm and quantitative analysis, the algorithm proposed in this paper with high accuracy. The deformation and shift exist in the prostate DWI sequence images caused in the process of imaging has effectively solved.