Research On Multispectral Bioluminescence Tomography Based On Primal-dual Active Set Algorithm

Bioluminescence tomography(BLT)is a new type of optical molecular imaging(OMI)modality,which can visually observe the labeled cells in the organism.It has broad application prospects in pre-clinical research such as understanding the development process of disease,early detection of malignant tumors,therapeutic drugs and quantitative evaluation of programs.However,the reconstruction of BLT is restoring the position and intensity of the light source inside the organism through the luminous flux information obtained by the surface measurement,which is a typical ill-posed problem.At the same time,the absorption and scattering of light by biological tissues causes the measurement is very limited,increasing the difficulty of numerical solution.In order to reduce the ill-posedness of BLT reconstruction,this paper combines the multispectral measurement method to obtain more luminous flux information.Besides,the paper launches a series of optical molecular tomography studies based on the primal dual active set(PDAS)algorithm,and respectively designed simulation and mouse in vivo experiments to verify the feasibility and effectiveness of the algorithm.The research content of this paper mainly includes:(1)Combining multi-spectral information,a unified non-convex regularization model for solving multi-spectral bioluminescence tomography is established,and the primal dual active set algorithm is introduced to solve different non-convex regularization models.A series of simulation experiments on the digital mouse proved the wide applicability of our proposed algorithm on non-convex regularization models.(2)In order to further improve the results of BLT reconstruction,a multi-spectral BLT reconstruction algorithm based on primal dual active set with continuation(PDASC)is proposed on the basis of PDAS,which combines the continuous technology of regularization parameters.And the optimal solution can be obtained by automatically to adjust the regularization parameters.The simulation results of the inhomogeneous digital mouse model show that the BLT algorithm based on the PDASC exhibits more accurate positioning errors and better shape-fitting capabilities in the single-light and dual-light experiments.The mouse experiment results also further verified the feasibility and potential of this method in practical applications such as tumor detection.(3)As a kind of structural prior information,group sparse information can group variables and judge the sparseness according to whether the variables are zero as a whole,which can better reconstruct the shape of the light source.On the basis of PDASC,combined with group sparse prior information,a multi-spectral BLT reconstruction algorithm based on group primal dual active set with continuation(GPDASC)is proposed.A series of simulation experiments proved that our proposed group sparse model not only performs better in large light source reconstruction,but also has stronger dual-object discrimination ability and higher shape fit compared with other group sparse algorithms.The mouse experiment also verified its potential in some practical applications.