Research On PET/CT And PET/MRI Registration Algorithms

Positron Emission Tomography(PET),as one of the most advanced medical imaging technologies,can provide information about human metabolic function at the molecular level and provide necessary basis for doctors’ clinical diagnosis and treatment.This paper mainly studies the registration algorithm between low-resolution PET images and different forms of medical images.The purpose is to provide data for PET attenuation correction and obtain anatomical and functional information of the lesion area.The main research content of this topic is divided into four parts:1.Image segmentation and enhancement are performed on the characteristics of low-resolution PET images such as noise,artifacts,and blur.The CV model head segmentation algorithm based on the morphological a priori shape can effectively segment PET head images with unclear edges,prominent areas and missing edges.The histogram matching method is used to enhance the PET image according to the specified histogram.The resulting PET image has higher edge contrast and clearer internal information.The preprocessing results provide high quality original images for subsequent experiments.2.Aiming at the registration study of PET images and different modal images,a registration scheme is proposed: A space transformation model combining affine transformation and B-spline surface,Mutual information as similarity measure,Improved Powell as a search algorithm.Experimental results show that the proposed algorithm achieves ideal registration effect and efficiency for low-resolution PET head images and images with different modes.In this paper,the structural similarity coefficient SSIM,which is used to simulate the quantitative evaluation of human eyes,is used to conduct quantitative evaluation on the registration results,and a judgment consistent with the evaluation of human eyes can be obtained.3.Aiming at the problem of narrow field of vision of single-mode multi-bed medical image,a splicing algorithm based on block matching combined with wavelet fusion is proposed.Block matching can quickly find the overlapping part of the image,and then perform wavelet fusion on this part to realize image splicing.Experimental results show that the proposed algorithm can accurately splice organs and bone positions without any artifacts or ghosting,and the splicing line is smooth.The stitching effect of blurred images and incomplete overlapping images can still be ensured,providing a single mode image with a comprehensive field of vision.4.Aiming at the registration method and splicing method proposed in this paper,MATLAB programming platform was used to design and realize the medical image registration operating system.Display the function of rewriting medical image header file,adjusting the window width of image window level,adjusting the image size,the image splicing function of single-mode multi-bed body image,and the image registration function of different modal heads of the system.