介入治疗导航系统的多模态图像配准方法分析-机械电子工程专业论文.docxVIP

哈尔滨工业大学工学硕士学位论文 哈尔滨工业大学工学硕士学位论文 - - II - Abstract Interventional therapy has been recognized as the most effective treatment for liver cancer because of its safety, efficiency, minimally invasive and speedy recovery. However, doctors cannot observe the specific situation of the lesion area in a minimally invasive surgery. As a result, the image- guided system is desiderated to provide the accurate position of lesion area to the doctor. Image registration is absolutely a key technology for surgical navigation system. Research has shown that registration can make great contributions to image fusion and target location directly or indirectly. And researchers have been dedicated to improve the accuracy, commonality and responsiveness of image registration for a long time. Firstly, the main principles of registration and registration framework are given based on the analysis and comparison of different image- guided surgical systems and the current research achievement in the field of mult i- modal image registration and 2D-3D registration in and abroad. Then implementation process is discussed and realized with MATLAB, laying a theoretical foundation for establishing a platform for researching image registration. Secondly, A research platform based on mutual information is given with four elements integrated into registration frame in the standard module, and the efficiency of rigid transformation, affine transformation, Demons algorithm and B- spline surface are analyzed as different transformation functions in ITK. Thirdly, a new hierarchical registration strategy, combined with LBFGS and gradient descent optimization algorithm, is proposed with different sample ratios, control points of B- spline surface, searching step sizes and convergence conditions. The computation efficiency and accuracy are improved. Finally, the 2D-3D image registration is achieved with the CT conversion and DRR technology. The hierarchical registration strategy and the traditio