Sparse Reconstruction Of Bioluminescence Tomography Based On Iterative Support Set Detection

Bioluminescence tomography(BLT)is an optical molecular imaging technique with high sensitivity and specificity,which can provide three dimension distribution of the internal source from the detected boundary light intensity.It can monitor organism physiological and pathological changes in the organization dynamic at the molecular level and can be used in some fields like the observation of tumor growth,the cell lesion detection and drug development.BLT,an ill-posedness inverse problem,is facing a huge challenge because of the complexity of the light transmission in the tissue and the distribution of limited measurements on the surface.Imaging algorithms become the key to apply BLT to the preclinical research.In order to get the more stable and accurate reconstruct result with limited measurements,this paper adopts the first-order Diffusion Approximation(DA)of the Radiative Transport Equation to describe the light propagation in biological tissue.Combined with the features of the sparse distribution of the light source in the application of BLT,this paper first formulates the BLT source reconstruction into a L1 norm minimization problem,then based on the Truncated Basis Pursuit(Truncated BP)model,this paper proposed a light source reconstruction algorithm based on Iterative Support Detection(ISD),which can reconstruct the bioluminescent source accurately rapidly and stably with less measurements.Simulation experiments in single source case mainly explore the influence on reconstruction with different organs,noises and threshold parameters,while simulation experiments in double source case mainly explore the ability of the algorithm resolving ability with different source distance.The results show that ISD has great performance with different optical parameter and noise level while the threshold parameter has some influence on the results,and it can also reconstruct the source accurately under different distance of the target.In order to verify the ISD algorithm further,this paper used the algorithm to reconstruct the sources compared with other five different algorithms.They are Tikhonov regularization algorithm,the Basis Pursuit(BP)algorithm,the Iterative reweighted least squares(IRLS)algorithm,Incomplete variables truncated conjugate gradient algorithm(IVTCG)algorithm and Stagewise orthogonal matching pursuit,(StOMP)algorithm.Simulation results show that the proposed BLT reconstruction algorithm can yield accurate reconstruction in both single-source and double-source case.It has great source localization ability and low time complexity.