2D-3D Medical Image Registration Research In Orthopedic Surgical Navigation System

Combining closely with the urgent need of orthopedic surgery navigation system for accurate image registration,this article,in order to solve the problems in 2D-3D registration such as the small range of parameter transformation and the difficulty of multi bone image registration,systematically carries out the research of medical image processing,the overall scheme design of 2D-3D registration,the construction and experiment of compilation environment.The main research achievements are as follows:1.Research on medical image processing methods,such as target extraction based on region growing and plane projection based on digital reconstruction image.Firstly,the imaging characteristics of two-dimensional intraoperative fluorescence images are analyzed,and the region of interest is extracted for the images with artifacts or multiple bone blocks to realize the segmentation and assignment of planar images.Secondly,the3 D data of the bone block is determined according to the target to be registered,the CT of the patient is extracted by Mimics software,and the target volume data is obtained by correlation processing.Finally,the spatial transformation model is selected according to the characteristics of registration target,and the DRR image algorithm is designed based on the principle of ray projection.It provides a data basis for the realization of 2D-3D image registration technology.2.Initial pose estimation and image registration are designed based on orthogonal fusion projection principle.Firstly,the image registration process is divided according to the 2D-3D registration link.Secondly,an initial pose estimation method is proposed for the initial search space pose parameters.By constructing the projection imaging model,the transformation parameters of the space rigid body are decomposed into the positive and lateral planes,and the mapping relationship between the space parameters and the plane parameters is established.Finally,the orthogonal biplane templates generated by projection are matched with the corresponding target image,and the registration parameters of the positive and lateral planes are converted into spatial parameters,so as to complete the effective estimation of the initial pose of the threedimensional object.It describes some registration methods,similarity measure,gray interpolation and optimization methods in the process of image registration,and analyzes some common methods.3.Experimental design and result analysis of 2D-3D image registration.According to the needs of image registration,we choose the implementation of compiling environment scheme,combined with the acquired CT data of patients,aiming at the skull,femur and fractured femur objects.we complete the verification of the image registration method in this paper,and compare it with the traditional 2D-3D image registration method.The experimental results show that the proposed method can make the initial transformation range of the five spatial parameters reach ±30mm or ±20°,and different research objects within this range show good initial pose estimation effect.Compared with the traditional image registration method,the overall registration time of 2D-3D iterative optimization based on the above method is 46.5% shorter on average,the registration accuracy of single translation parameter is up to 69.2% at highest,and the registration accuracy of single rotation parameter is up to 91.1% at highest.At the same time,the final registration results also have good robustness.In this paper,starting from the practical problems of 2D-3D image registration,we can achieve the registration of single object,single bone block and multiple bone block images under large initial pose deviation,which will help to further improve the accurate and minimally invasive operation of orthopedic surgery navigation system,reduce the burden of learning and using for clinicians,and provide the best or near the best search parameters for multimode image navigation system.