Research On Registration And Fusion Of Multi-modal Medical Images In Neurosurgery

With the rapid development of minimally invasive surgery,contemporary minimally invasive surgery has gradually replaced open surgery.Minimally invasive surgery uses specialized surgical instruments to greatly reduce the surgical incision,thereby reducing the incidence of injury and postoperative complications,and greatly shortening the hospitalization and recovery time of patients.However,this surgical technique is difficult,high-risk,and requires doctors to have extensive experience.At present,only a few hospitals can perform it.There are three difficulties in performing this operation: first,the surgical field is limited,and the doctor cannot clearly observe the relative position of the brain in the lesion;second,the lack of positioning and navigation information of the surgical instrument increases the risk of the operation;again,heavy dependence During the CT image of the operation,the accumulation of radiation caused serious damage to the doctor’s body.To solve the above problems,people try to use new technologies.Among them,image-guided computer navigation technology is currently one of the technologies that are generally concerned.This article focuses on multimodal medical images of neurosurgery.The specific research contents are as follows:1.Aiming at the difficulty of localization of minimally invasive surgery,an image-based intraoperative navigation system was studied.The principles and main uses of medical image imaging in different modes were introduced.Summary and analysis of the development of fusion methods.2.This paper studies the image preprocessing to be completed before medical image registration,and introduces the importance of region of interest segmentation before medical image registration,the research status and common algorithms of image segmentation.Perform targeted segmentation of regions of interest in multimodal medical images and perform experimental verification.3.Aiming at the problem of registration accuracy of sequential medical images,two-dimensional image registration and three-dimensional point cloud registration are studied.This paper analyzes the principles and applicable scopes of the most commonly used registration algorithms in recent years,and improves the existing Super-4PCS algorithm.An improved Super-4PCS point cloud registration algorithm based on RANSAC algorithm is proposed.4.This paper researches on medical image fusion,one of the cores of image-based intraoperative navigation systems,and proposes a new three-dimensional fusion method for the problems existing in the current medical image fusion process.Different from common three-dimensional fusion,the three-dimensional fusion method proposed in this paper is to first three-dimensionally model the sequence medical images,and then fuse the model to effectively avoid the problems of angle and slice thickness differences in the two-dimensional sequence registration.5.For the medical registration and fusion algorithms studied in this article,experiments were performed on CT and MRI sequence images.In order to prove the applicability and accuracy of the registration algorithm,two models in the 3D point cloud database and dogs with labeled points were simultaneously The leg bone images were experimentally verified.The above experimental data uses the improved registration algorithm proposed in this paper to make the accuracy reach millimeter level.At the same time,by using the new three-dimensional fusion algorithm proposed in this paper for experiments,the fusion effect has also been affirmed by the doctors of the partner hospitals.6.This article summarizes the effects and applicability of the multimodal medical image registration and fusion algorithms applied in the image-based intraoperative navigation system,and looks forward to the future development of the field.