Research On Image Reconstruction In Freehand 3D Ultrasound Imaging System Based On Optical Positioning

With the increasing use of medical three-dimensional(3D)ultrasound images in clinical disease diagnosis and computer-assisted guided interventional surgery,3D ultrasound imaging has been extensively researched.The optical positioning and tracking device is widely used in freehand 3D ultrasound imaging system because of its advantages of high positioning distance and accuracy.In this thesis,the composition and function of the freehand 3D ultrasound imaging system based on optical positioning are analyzed around the requirement of freehand 3D ultrasound high-precision imaging,and the research on the 3D image reconstruction process of the system is completed.We propose the effective solutions by analyzing the existing problems of the image reconstruction algorithms and the process of freehand 3D ultrasound image reconstruction.The main works of this thesis are as follows:(1)The process of 3D ultrasound image reconstruction in the freehand 3D ultrasound imaging system based on optical positioning is analyzed,and the key technologies in image reconstruction are researched,such as the extraction of image interest region,the construction of voxel data bounding box,the reassignment of pixel values and the visualization of 3D image.(2)According to the problem existing in the process of 3D ultrasound image reconstruction,in this thesis,a function module for denoising is added to the process of 3D ultrasound image reconstruction.Based on the advantages and disadvantages of the existing ultrasound images denoising algorithms,in this thesis,we adopt extreme learning machine algorithm to remove the noises in image.The principle of the algorithm is analyzed,and the denoising parameters of the algorithm are determined by experiments,and the denoising process of the algorithm is derived.Finally the denoising performance of the extreme learning machine algorithm is tested on ultrasound image.The results show that extreme learning machine algorithm has a good denoising effect and also protects the edge of images.(3)We take a comprehensive analysis on the existing 3D ultrasound image reconstruction algorithms.According to the problems of these existing algorithms,in this thesis,the adaptive kernel regression algorithm is applied to the new field of 3D ultrasound image reconstruction based on kernel regression.The principle of the algorithm is analyzed,the calculation of adaptive regression kernel and the process of 3D ultrasound image reconstruction by adaptive kernel regression algorithm are deduced.Finally the reconstruction performance of the proposed algorithm is tested on tumor ultrasound image.The results show that the proposed algorithm has a good denoising effect and protects the edge of image in 3D ultrasound image reconstruction.As the accuracy of the proposed algorithm is high,it is helpful for segmentation and recognition in image post-processing.(4)Based on the two presented algorithms in this thesis,the test of 3D image reconstruction on tumor preoperative and postoperation ultrasound images are completed.By the qualitative and quantitative analysis of the reconstruction results,we know that the adaptive kernel regression algorithm is a kind of 3D ultrasound image reconstruction method with high reconstruction accuracy,and it is helpful to improve the accuracy of 3D image reconstruction by taking a denoising operation on ultrasound images.Lastly the feasibility and effectiveness of the proposed algorithms are verified.