Medical Image Registration Based On Volumetric Feature Points And Maximum Mutual Information And Its Application In Bone Tumor Surgery

Image registration is a key technology in the process of obtaining comprehensive information about disease from different medical images.Image registration is imposing a kind of spatial transformation of floating image,which makes the floating image and the reference image pixels in the spatial position to achieve the same.In the preoperative planning of patients with bone tumors,the spatial location of bone tumors in human tissues can obtained from CT images,and the size of bone tumors could be precisely defined from MRI images.The registration of CT images and MRI images can accurately define the spatial distribution in human tissues and size of bone tumors,which can greatly improve the accuracy of bone tumor surgery.How to achieve accurate registration between different images is the subject of this paperIn order to meet the requirements of medical image registration,this paper proposes a medical image registration method based on the feature points of the human and the maximum mutual information which divided the image registration into two parts: the first registration and the second registration.By matching the volumetric feature points based on volume visualization and searching the maximum mutual information to realize the goal of image registration.The main contents of this paper are as follows:1)In the process of 2D image registration,the normalized mutual information is used as the similarity measure of registration,and the method of searching the maximum mutual information parameter based on improved genetic algorithm and improved Powell algorithm is proposed.The roughness result of the improved genetic algorithm is used as the initial point of the improved Powell algorithm,and finally realized the image registration based on the maximum mutual information.2)In the process of image registration,as the floating image and the reference image pixel spacing and layer spacing is different,making the deformation in each direction is inconsistent in the process of image transformation.In this paper,a method of data field normalization is proposed in image registration.Accurately determine the direction of the zoom factor,making the process of image registration more accurate and fast.3)In this paper,we adopt the method of visualization technology based on surface rendering.Volumetric visualization is mainly used in the selection of the volumetric feature points,visualization of image registration results,and resection cases in actual bone tumors.In addition,a method of multi-sequence images surface rendering is proposed on the basis of single-sequence images surface rendering,that's mainly designed for the process of image registration which is needed to import multi-sequence images.4)In the process of 3D image registration,a new image registration method based on the volumetric feature points and the maximum mutual information is adopted.The initial image registration is to select the volumetric feature points interactively and make them aligned based on the Volume visualization,to achieve the original data field space alignment purposes.Secondary images registration is the process of accurate registration of images based on maximum mutual information after initial registration.5)Image registration is applied to the actual femoral tumor resection,the tumor model and the femur model were obtained by three-dimensional reconstruction.After the bone tumor model and the femur model were registered,the spatial location and size of the tumor were accurately determined,which will improve the accuracy of tumor resection.The experimental results show that the image registration based on the volumetric feature point and the maximum mutual information can achieve high accuracy,which has important reference value in the practical application of clinical practice.