Reconstruction Method Research On Fluorescent Molecular Tomography Based On Three-decision-making

In Fluorescence molecular tomography?FMT?,according to the diffuse light distribution detected at the boundary,an appropriate reconstruction method is selected to reconstruct the three-dimensional distribution and fluorescence concentration of the fluorescent target in vivo.As a result,the physiological and pathological processes of organisms can be monitored at the molecular level.Due to high sensitivity,low cost,and non-invasive advantages in FMT,it has become an important branch of molecular imaging and has been successfully used in cancer diagnosis,drug development,and therapeutic evaluation.However,because the absorption effect of light propagation inside the tissue is much lower than the scattering effect,and the amount of fluorescence information on the biological surface obtained by the CCD camera is small,FMT has a strong ill-posed.So how to accurately reconstruct Fluorescent targets becomes the core issue of FMT.In this paper,we studied fluorescence molecular tomography reconstruction.In order to improve the stability of reconstruction,the feasible region is extracted based on the three decision-making theories.In the feasible region,the smooth l₀ norm is used to achieve the finite projection FMT reconstruction,and the TV norm minimization algorithm of the split Bregman is used to enhance the reconstruction image quality.1)Finite Projection FMT Reconstruction Based on Three Decision Theory and Smooth l₀Norm.In Finite projection fluorescence tomography?FMT?,three-dimensional distribution of fluorescent targets in vivo can be quickly reconstructed with a short data acquisition time.However,the limited projection FMT has serious morbidity due to less projection data.In order to reduce the morbidity of the finite projection FMT and improve the reconstruction accuracy,and considering the characteristics of sparse distribution of light sources in FMT,a finite projection FMT reconstruction method combining three branch decision theory and smooth l₀ norm is proposed.In this method,a smooth l₀-norm-based FMT reconstruction algorithm is used to solve the reconstruction.It uses a similar continuous function to approximate the l₀ norm.It can achieve fast solution.At the same time,the three decision-making theories are applied to extract the feasible region during reconstruction,and the algorithm is used to perform multiple reconstructions on feasible regions improving reconstruction accuracy.The digital mouse model reconstruction results show that the position error of the reconstructed images is less than 1mm at 3,6 and 9excitation points,and the reconstruction time is short.The reconstructed results of the physical experiments further demonstrate the effectiveness of the proposed method in actual FMT reconstruction.2)FMT reconstruction based on three decision theory and TV norm regularization.The l₁norm regularization and l₂ norm regularization methods have been widely used to solve fluorescence molecular tomography problems.However,using regularization of the l₂norm will lead to the over-smoothness of the reconstructed fluorescence target.Using Regularization of the l₁ norm may sparse the distribution of fluorescent light sources and further reduce the quality of the reconstructed image.In order to improve the reconstructed image quality,FMT reconstruction is equivalent to the TV norm minimization problem,and non-negative constraints are added in FMT.Adding an auxiliary variable by splitting Bregman method,and the TV regularization optimization problem with non-negative constraints is equivalent to two independent sub-optimization problems that can be solved by the existing iterative algorithm.At the same time,the three decision-making theories are applied to extract the feasible region during reconstruction to improve the stability of FMT reconstruction.A simulation experiment was set up to verify the recovery ability and stability of this method for single light source and dual light source,and its practicality was verified by physical experiments.The experimental results show that the combination of three decision-making theories and TV norm regularization method can accurately and stably reconstruct the fluorescence target distribution and improve the image quality at the same time.