Research On Segmentation And Registration Technology In Medical Image Guidance

In recent years,thanks to the continuous rapid development of computer science technology and electronic imaging technology,it has provided convenient conditions and scientific basis for medical imaging to guide clinical surgery,as well as many significant technical challenges.The key part of segmentation and registration in medical image-guided clinical surgery technology has also become a research hotspot and difficulty in the field of computer-assisted hospital surgery.This article takes the application of medical image-guided clinical surgery as the background,takes the knee joint medical image as the research object,and focuses on the exploration and research of segmentation and registration in the technical process.For the segmentation of knee joint images with adhesion or overlap problems,design and propose related improved segmentation algorithms,and design and propose related registration algorithms for the registration of 2D images and 3D images of knee joints.The main research work of this paper is as follows:First of all,in order to achieve the segmentation of the region of interest in the knee joint image,the existing segmentation algorithm is studied.Aiming at the overlap and adhesion problems,a semi-automatic segmentation algorithm based on region growth and level set joint segmentation is proposed.First analyze the DRLSE algorithm model,redefine the edge stop function,set the evolution adjustment factor,and then adjust the size of the factor according to the image gradient information to make it have a smaller function value at the edge overlap and adhesion position,so as to get the correct evolution result;Finally,combined with the region growth,the outer boundary with thick edges is segmented to obtain a complete segmented image.Through experimental comparison,the feasibility of the algorithm in this paper is proved.It can effectively segment the adhesion and overlap in the knee joint image,and can also correctly identify the edge of the noise image.The algorithm has good robustness.Secondly,the medical image registration technology is researched,verified and improved.For the registration of 2D images and 3D images of the knee joint,a combination of algorithms is proposed based on the existing registration algorithms: In the standard,the gradient angle is introduced as the similarity measurement function,and the CMA-ES algorithm is introduced as the parameter optimization method.In terms of algorithm implementation,GPUs supporting the CUDA architecture are used for acceleration to complete rapid registration;Experimental comparisons prove that the registration algorithm combination in this paper has stronger advantages than traditional methods in terms of registration speed and accuracy.Finally,using VS2017 and the VTK open source library as the development basis,a simple medical image processing system is designed and implemented.The main function is to realize the reading of medical image slices,the reconstruction and visualization of 3D images.