Research On Medical Image Reconstruction Method Based On Norm Sparse Representation

With the significant improvement of modern living standards and the gradual advent of the post-epidemic era,the number of physical health examinations in medical institutions has shown explosive growth.Scanning analysis and transmission preservation of suspected lesions has become a hot topic in medical research.Among them,computed tomography and magnetic resonance imaging play an important role in the field of clinical medicine.However,they are faced with the problems that the ionizing radiation is proportional to the imaging time,and the imaging quality is affected by the scanning time respectively.In order to reduce the radiation and waiting time of patients during the examination,and improve the transmission and preservation efficiency of imaging information,the under-sampling recovery in compressed sensing theory is introduced.In this context,based on the shortcomings of existing research methods,this paper conducts in-depth research on the rapid reconstruction of two-dimensional static computed tomography images and magnetic resonance images based on smooth L0 norm,and the real-time reconstruction of threedimensional dynamic magnetic resonance imaging based on L1 norm and L2 norm.The specific research contents and results are as follows.Firstly,this paper studies the methods and characteristics of magnetic resonance imaging in medical images,explores the development process and contents of compressed sensing theory,and collects the data sets needed for simulation experiments.At the same time,this paper also verifies several classical signal reconstruction algorithms based on compressed sensing,selects appropriate and credible reconstruction performance evaluation indicators,and draws the necessary conclusions of the research.Secondly,aiming at the problems that are difficult to solve and unable to balance image quality and recovery time in two-dimensional magnetic resonance imaging based on compressed sensing,a two-dimensional magnetic resonance image reconstruction algorithm(CCGSL0)based on composite cosine function family and conjugate gradient descent method is designed and implemented.The composite cosine function family fits the L0 norm better and can quickly find the global optimal solution of the objective function.At the same time,the block processing mechanism and the conjugate gradient method can highlight the local information characteristics of the sub-block image and improve the convergence speed of the algorithm,thereby reducing the reconstruction time of the overall operation.Compared with the classical efficient reconstruction algorithm,CCGSL0 has better two-dimensional static medical image restoration performance.Finally,aiming at the problems of excessive multi-frame reconstruction time and inability to achieve real-time recovery in 3D magnetic resonance imaging based on compressed sensing,a 3D magnetic resonance image reconstruction algorithm(L12-SSR)based on weighted dynamic total variation operator and parallel double-tree complex wavelet transform is designed and implemented.The weighted dynamic total variation operator can solve the time domain redundant operation of the signal,reduce the error influence in the operation process,and initially improve the reconstruction accuracy and speed of the signal.The parallel dual-tree complex wavelet transform can solve the spatial redundancy operation of the signal and further reduce the signal reconstruction time.At the same time,the SplitBregman algorithm is used to simplify the problem to be solved.Compared with other dynamic magnetic resonance image reconstruction algorithms,L12-SSR accelerates the speed of signal reconstruction and can realize approximate real-time recovery of dynamic magnetic resonance images,thus saving the storage space and time of three-dimensional images in medical research.