Research On Medical Image Registration With Applications In PET-CT System Of Radiation Therapy

As the development of medical imaging techniques, they have been widely applied to clinical diagnosis and treatment, and there are multi medical imaging modalities in practical application, these imaging techniques can be divided into two types:anatomical and functional imaging techniques. Anatomical imaging provides structural information images with high-resolution, and functional imaging provides human body functional metabolism information images with coarser resolution. PET-CT system is a complementary imaging technology with the dual advantages of PET and CT images, combined whole-body anatomical structure and functional metabolic information. Image registration techniques can make the multimodality image complementary information fuse better, providing richer clinical information, and doctors can obtain the patients’ accurate information, in order to improve the diagnosis and treatment of clinical disease.In recent years, medical image registration technology is one of the important branches in the field of medical image processing, and is the key technology of preprocessing in image fusion and multimodality image analysis. With the advance and application of digital image processing and computer technology, the medical image registration technology has been developing rapidly. According to the demands of clinical application, the accuracy and speed of medical image registration technique are needed to improve, to meet the practical requirements of quickly and accurately positioning the target tumor, and reducing the setup errors.In this paper, PET-CT multimodality imaging system is treated as research object, and we improve and solve some problems existing in multimodality imaging technology, and deeply research the registration algorithm of multimodality images. The proposed and improved algorithm are applied to clinical tumor radiotherapy, precise positioning the tumor target of radiotherapy treatment planning system, improving the patient’s radiotherapy effect on the basis of protect the normal tissues and organs.Demons algorithm, based on optical flow field, has the advantages with fast process speed and high accuracy, but this method drive images to transform only depends on grayscale difference and grayscale gradient images, so this method is only suitable for single-modality image registration. In this paper, based on the traditional Demons algorithm, we adding an additional external force based on the gradient of mutual information between two images for current transformation, which can achieve PET and CT images deformable registration. The experimental results show that, the Demons algorithm based on mutual information gradient can accurate positioning the tumor target, and improves the precise and effectiveness of radiation therapy.Based on B-spline FFD registration method has high accuracy, but with slower process speed. This paper combines the features of tumor image, the setup errors are in the overall image, and deformation errors caused by breathing and heartbeat are in the tumor area and surrounding normal tissues and organs. Therefore, the large-spaced grid is used in the overall image, and global rigid registration algorithm is used to minimizing the setup errors; the small-spaced fine grid is used in the internal tumor region, and using B-spline-based FFD method. This based on non-uniform grid deformation registration algorithm not only ensure the registration accuracy of tumor image, but also improve the registration speed.