Research On Brain Medical Image Registration Algorithm Based On Features

As an important method of disease diagnosis,medical imaging has attracted increasing attention.In particular,medical brain imaging is often used as an important basis for the diagnosis and treatment of brain diseases,and it has become one of the current research hotspots in computer vision.The importance of brain medical image registration technology as the basis for complex tasks such as medical image fusion and target detection is self-evident.Among different methods,the feature-based brain medical image registration technology has significant advantages such as fast speed,strong robustness,and good pertinence.However,the accuracy of this method is slightly worse than that of medical image registration based on full pixels,and it is greatly affected by the features of objects in the image and environmental noise.Therefore,this article focuses on the above proposed methods.The main content and innovation work are summarized as follows:1)This paper summarizes the principles,advantages and disadvantages of common medical imaging,registration classification methods,basic processes,and the required key technologies,and introduce the important modules in the medical image registration process in details,including image similarity measurement algorithms,spatial transformation techniques,image interpolation methods and the used search optimization algorithms in the registration process.2)Aiming at the registration method based on feature point description and according to spider local image feature(SLIF)sampling model which has the advantages of wide sampling range(compared to the traditional sampling model which has the same number of sampling points)and adjustable sampling range,a new description is designed in this paper,which is called improved spider local image feature(ISLIF)descriptor.This descriptor uses the self-similarity of the image as the description basis,and takes advantage of the standard deviation as the adaptive threshold to adjust the similarity to describe the similarity characteristics of the image;at the same time,the description of the spatial structure information of the description is enhanced.The global neighborhood,circular neighborhood,and radial neighborhood in the sampling range are used as the description content and included in the feature descriptor which follows the principle that desgining feature descriptors need to take the features of the image immutable and distinguishable into consideration,and greatly improves the robustness and stability of the descriptors.The proposed method brings a significant improvement in the subsequent medical image registration performance.3)Aiming at the registration method based on image segmentation,this paper uses the bias corrected fuzzy c-means algorithm(BCFCM)image segmentation method to process multi-modal MR images.This algorithm can achieve a good segmentation effect for images with uneven illumination.And then,combined with binarization threshold segmentation,MSR symmetry axis detection algorithm and the proposed sparse symmetry or dense symmetry constraint operator,the multi-modal MR image is finally converted into an artificial modal image for registration,and its registration accuracy is greatly improved.The proposed method makes good use of the consistency information of the modal images(the brain is approximately symmetrical)to constrain the registration of multi-modal brain images,and it combines image segmentation and symmetric algorithms for image registration algorithms.At the same time,this paper also puts forward some reference suggestions on the processing effect of the symmetry constraint operator proposed in the method and the free variables involved in the method combined with related experiments.The proposed algorithm in this paper will be performed on synthetic medical images and real patient images.In order to facilitate comparison,comparative experiments of related methods are also introduced.Experiments prove that the proposed algorithm has significant advantages in anti-noise ability and registration accuracy.