A Simulation Study On The Shape Reconstruction Of The Light Source For Multispectral Bioluminescence Tomography

As a new imaging method which emerged in the early 21st century,bioluminescence tomography is an important imaging modality of optical molecular imaging.Different from fluorescence molecular tomography,BLT makes use of bioluminescence of target cells through transgenic technology and doesn't require external light source excitation.BLT will not be interfered by the auto-fluorescence from other tissues.It has low imaging cost and high signal-to-noise ratio.Therefore,it has great potential in the early detection of cancer,research and development of new drugs,image-guided radiotherapy and other pre-clinical research.When bioluminescence tomography is applied to guide the radiotherapy for cancer,the position of the light source corresponds to the distribution of the target cells,the size and shape of the light source is the size and shape of the target(such as tumor tissues).They directly affect the choice of radiotherapy area.Therefore,the shape of the reconstructed light source is very important for such applications.However,due to the absence of external light excitation in bioluminescence tomography,the fluorescence of near infrared spectra undergoes complex absorption,scattering and other transmission processes in organisms.Consequently,the signals detected on the surface of organisms are weak,which results in the measured data are seriously insufficient.How to deal with the serious ill-posedness of the inverse problem of BLT and how to stably and accurately reconstruct the spatial position,shape and size of internal target using limited measurement information are a challenging problem of BLT.To address the above problem,the main research of this thesis includes:(1)Firstly,the influence of the light translission model on the size and shape of the reconstruct light source are investigated.Specifically,the third-order simplified spherical harmonics(SP3)of the Radiation Transfer Equation(RTE)and the traditional diffusion approximation(DA)method are compared in this thesis.The ratio of the shape fitting degree of the two reconstruction methods is quantitatively evaluated by the Ratio between the overlapping and total volume(OVTVR)of the reconstructed light source and the real light source.The SP3 model with better performance was selected for further research.(2)Based on the SP3 model,the light source is reconstructed with the incomplete variable truncated conjugate gradient(IVTCG)algorithm,and multi-spectral measurement data are used to increase prior information and reduce the ill-posedness of the inverse problem.The simulation experiments on several groups of digital mouse suggest that the reconstruction method achieves better results in different positions,sizes and even optical parameters,but this scheme is very time consuming.(3)For the purpose of accelerating the reconstruction process,we propose an improved reconstruction method,i.e.,using the non-convex L1/2 regularization iterative reweighted-L1/2(IRW-L1/2)algorithm combined with the SP3 model,which reduces the time cost of light source shape reconstruction and produces better reconstruction results.The simulation experiments on the digital mouse takes into account the position,shape,size and optical parameters of the light soxurce.The experimental results show that the reconstruction method achieves similar or better reconstruction results with lower time cost.