Research On Gradient-based Deformation Registration Algorithm Combined With A Rigid Penalty Term

Image guided radiation therapy(IGRT) has been widely used in clinical radiotherapy. It is mainly used to solve the pendulum position error between clinical radiotherapy times, the shrinkage of target tumor, respiratory motion during radiotherapy and other unfavorable factors having some adverse effects on radiotherapy accuracy. Medical image processing technology is the key technology in IGRT system, including image registration, image segmentation and image fusion and so on. The application of medical image technology to solve the adverse factors of clinical radiotherapy, radiotherapy planning and optimization, improve the accuracy of adaptive radiotherapy is the hotspot of many experts and scholars at home and abroad.This paper mainly introduces the development background, research status and significance of image guided radiotherapy, and the process of image registration. A rigid penalty term is proposed for non-rigid deformable registration, which is used to constrain the transformation of rigid body in non-rigid registration. At the same time, the deformation registration based on gradient, we will combine the rigid penalty and gradient-based registration algorithm, and make a contrast with the former algorithm, analyzes the advantages of our proposed new algorithm and its significance on clinical adaptive radiotherapy.In this paper, the focus is the application of a combination of the rigid penalty term and gradient-based deformable registration algorithm in adaptive radiotherapy. So we introduce a rigid penalty term and research process of gradient-based deformation registration algorithm, which mainly contains the following research work:(1)Research on the rigid penalty term. In medical image, there are only two kinds of structures, the rigid structure and the deformation structure. Some deformable registration algorithm that can’t simulate structure characters of different types of organization structure may lead to rigid structure deformation. In this paper, a local rigid penalty term is designed, to prevent the deformation of rigid structure, such as skeletal. Through the gradient vector to represent the deformation field, a fast algorithm was designed, compared with the unconstrained registration. That was evaluated in clinical chest CT data and the result shows that the rigid deformation registration combined with the penalty term can keep the accuracy of deformation registration and maintain, the local rigid structure.(2)In adaptive radiation therapy, the registration of Computed Tomography(CT) and Cone Beam Computed Tomography(CBCT) is also a crucial step in radiotherapy. Due to the electronic density between CT and CBCT images is inconsistent, the intensity-based registration algorithm is not suitable for the registration of CT and CBCT. Gradient-based deformation registration algorithm, using gradient vector field to describe the deformation field, can reduce the influence of CBCT intensity inaccuracy for similarity measure. The local polynomial approximation-intersection of confidence intervals(LPA-ICI) algorithm was used to describe smoothing process for constructing the model to ensure a smooth and continuous deformation field. A‘bi-directional’ force was introduced to accelerate the registration process, at the same time multiscale strategy was used to speed up the registration process to avoid the local extremum.(3)The new algorithm is a gradient-based deformation registration algorithm combined with a rigid penalty term. It not only solves the problem of deformable registration accuracy between CT and CBCT as different intensity, while it also completes the rigid registration of rigid structure in non-rigid registration, to improve the accuraxcy of the adaptive radiotherapy.