Registration Assisting System For 3D Positioning Based On 2D Dual-fluoroscopy Medical Images

Dual Fluoroscopic Imaging System(DFIS)is one of the main application tools for clinical diagnosis in the field of sports medicine.The dynamic X-ray image data collected by it can be used for 2D-3D registration for sports Analysis and further research or diagnosis.In the analysis process of the existing DFIS data,the artificial solution has many shortcomings,such as the inability to quickly and accurately determine the bone contour,the large amount,highly repetition,and the lack of accurately of three-dimensional registration data,etc.To solve the above problems,this paper designs a dual fluoroscopic medical fluoroscopy image registration assisting system based on the knee joint image data collected by DFIS,which is used to assist medical workers to improve the speed and accuracy of the registration work.The main contributions of this article are:1.By studying the basic framework of existing medical image registration and according to the manual registration process,the dual fluoroscopic registration problem was modeled,and the problem was decomposed into three sub-problems of image noise reduction,image segmentation and parameter optimization.The main process and feasibility of using the registration system to solve the dual fluoroscopic registration problem was clarified.Due to the complexity of the actual registration problem,the assisting nature of the system is established,that is,the ability of the user to intervene and modify the results in each process of registration is fully given.2.To solve the image denoising and parameter optimization problems involved in the registration problem,modeling and analysis are carried out,this paper compared and analyzed the existing algorithms in those research fields.As the result,the total variation algorithm is selected as the system's denoising algorithm based on the actual engineering,BFGS algorithm is used as the optimization algorithm of the system.3.To solve the problem that it is difficult to use existing algorithms for automatic contour extraction of medical images acquired in actual work,a local strategy based on gray-scale similarity rules is proposed to control the growth of contours according to the step length.The contour is gradually generated,and the user can also manually intervene in the growth direction of the contour.4.Requirement analysis and architecture design of the registration assisting system were carried out,and the modules and operation process of the system were clarified.The system is implemented in C++ language programming based on Qt and VTK,and completes the functions of medical image analysis and unpacking,denoising preprocessing,interactive contour extraction,manual registration and automatic registration.The experiment proved the functional integrity of the system and verified that the registration assisting system effectively improves the speed and accuracy of the dual fluoroscopic image registration,and can assist medical workers in shortening the registration time.