Research On Reconstruction Methods For Fluorescence Molecular Tomography Based On Group Sparsity Prior

Molecular imaging(MI),as an emerging specific imaging technique,combined with modern molecular biology techniques,which can noninvasive visualize the dynamic molecular and cellular processes taking place in vivo in real-time.It has been widely used in disease diagnosis and treatment,drug research and surgical navigation.Optical molecular imaging technology,which uses light as the imaging medium,has made rapid development in the past decade due to its advantages of safety,non-radiation,low cost and high sensitivity.Fluorescence Molecular Tomography(FMT)is an important part of optical molecular imaging,which can be used for three-dimensional observation of specific cells and molecules in organisms.In recent years,although FMT has developed rapidly in reconstruction method and application,there are still some problems in FMT.The research of this thesis focuses on the major topic to improve the performance of FMT,including the accuracy,efficiency and morphology.First,we proposed to utilize the group sparsity prior of the tumors,and then,the L2,1-norm method and fused LASSO method(FLM)based on the group sparse prior were proposed.On the other hand,to further evaluate the performance of group sparsity,we proposed a reconstruction method which is primal accelerated proximal gradient based on L1-norm,and compared it to the group sparsity method to verify the performance of FMT reconstruction.The main contributions are listed as follows:1.We proposed to take group sparsity of the fluorescent signals as a priori information.In FMT,fluorescent probes cluster together in the regions of the tumors and that only the region of tumors has fluorescent signals.We assumed that the fluorescent signals in tumors were simultaneously non-zero and that normal tissues were simultaneously zero.Therefore,besides sparsity,fluorescent sources were also characterized by group sparsity.Group sparsity was widely used in compressive sensing,image denoising and signal processing.It has been proved that exploiting group sparsity has yielded more satisfactory results in morphology compared to using sparsity alone in many information processing studies.2.We proposed the L2,1-norm method based on group sparsity.The L2,1-norm is the sum of the L2-norms of the blocks of coefficients associated with each component.Because the L2,1-norm combines the advantages of L1-regularization and L2-regularization,the reconstruction of the L2,1-norm is neither over-smoothed nor over-shrunk.Thus,L2,1-norm is an effective reconstruction strategy for FMT.Comprehensive numerical simulations and in vivo experiments show that our proposed method was more accurate,efficient,and robust for fluorescence reconstruction compared to the Tikhonov-L2 method and IS_L1 method.3.We proposed the fused LASSO method(FLIM)based on group sparsity.The FLM based on group sparsity prior not only takes advantage of the sparsity of the fluorescent sources,but also utilizes the structure of the sources,thus making the reconstruction results more accuracy and morphologically similar to the sources..Both heterogeneous numerical simulation experiments and in vivo liver tumor mouse model experiments were carried out to verify the method.The simulated results proved the superiority of FLM over conventional methods in tumor detection and tumor morphological reconstruction.Furthermore,in vivo experiments also demonstrated that FLM offered the best performance in quantitative tumor imaging.4.We proposed a reconstruction method based on the sparsity prior,and utilized the primal accelerated proximal gradient(PAPG)for solving the object function.The simulation,phantom and in vivo experiments revealed that our method held advantages of robustness,accuracy,and efficiency in FMT reconstructions.However,compared to L2,1-norm method based on group sparsity,the performance of PAPG was not as good as L2,1-norm method,which further verified the advatages of group sparsity.