Research Of Registration Algorithm And Segmentation For Infant MR Images

The incidence of infantile cerebral palsy disease is becoming higher and higher.At present, the research of computer aided diagnosis of infantile cerebral palsy disease are still on the stage of theoretical research.With rapid development of computer technology, computer aided diagnosis method has become an important research field in a medical imaging, diagnostic radiology, computer science.Image segmatation plays an important role in medical analysis.Medical image segmentation aims to divide an entire image into different non-overlapping regions. Image registration is a good technique to fuse medical priori knowledge into segmentation procedure, which forms a new image segmentation, registration based segmentation. In registration based segmentation, an optimal spatial transformation deforming the source image into the target image is first obtained. Then the labeled structures in the atlas are mapped to the target image based on the obtained optical spatial transformation. Finally, the equired segmentation will be acquired easily.Firstly, this thesis studys the rigid registration algorithm of medical image.the methods proposed in the thesis treat mutual strictly concave function measure as the similarity measure,and PV or 3 orders B-spline GPVE for interpolation function are used for rigid registration for medical image. The study of rigid medical image registration algorithm solve the MR image rotation and translation problems due to the infant’s head caused by twisting, Then we study the diffusion model based Demons non-rigid registration algorithm in depth. The main researches focus on the intensity an spatial normalization,regularization method of the defomation field, the contributions of multi-resolution strategy to non-rigid registration.In order to decide the right parameter of driving force,the algorithm takes account of the influence of deformation to both gray and gradient at the same time while processing the problems of largely deformed image registration. At last, we get pediatric brain tissue and brain structure segmentation by using registration based segmentation and we contrast with traditional algorithm from qualitative analysis and quantitative analysis.Through experiments testing, the algorithm of infant brain image segmentation registration based segmentation in this paper can fast an accurate sement the tisse and deep brain structures. The algorithm of this thesis has greatly improved in segmentation efficiency and segmentation result, and has practical application value.