Research On Fast Reconstruction Method Of Mesoscopic Fluorescence Probe Spatial Distribution

Mesoscopic fluorescence molecular tomography（MFMT）,as a new and promising non-contact imaging modality,can perform high-resolution imaging at the molecular level,achieve early tumor screening,and provide important reference value for pre-clinical appli-cations.MFMT uses a highly sensitive near-infrared photon detection instrument to collect fluorescence information on the surface of biological tissues,combines with an accurate forward model to simulate the sensitivity matrix containing tissue optical parameters,and realizes the imaging and reconstruction of the spatial distribution of fluorescent probes in tumor lesions through three-dimensional reconstruction technology.However,MFMT reconstruction is a typical inverse solution problem.Highly ill-conditioned,ill-posed and large-scale numberical caculation problerms are the difficulties to be solved urgently.In order to solve the above difficulties,this paper focuses on the simultaneous algebraic reconstruction algorithm and fast iterative shrinkage-thresholding algorithm.The main research contents are as follows:（1）Aiming at the large-scale numerical calculation problem of mesoscopic fluores-cence molecular tomography reconstruction,a double pre-processing method based on prin-cipal component analysis is designed,and the large-scale sensitivity matrix used in the re-construction is processed in the order of first row and then column.At the same time,in order to maintain the consistency between the reconstruction result and the target data,a zero-padding method is used to restore the dimension-reduced matrix to the original dimen-sion.Compared with the conventional principal component analysis（PCA）,this method can significantly improve the reconstruction efficiency,has the characteristics of simple op-eration,flexibility and speed,and is feasible in large-scale data processing.（2）Aiming at the problem of sensitivity to noise caused by the high ill-condition of the inverse problem of MFMT,a rapid 3D reconstruction algorithm based on simultane-ous algebraic total variation is proposed.This method combines the simultaneous algebraic reconstruction technique（SART）with the three-dimensional total variation（TV）regular-ization method solved by Split Bregman,which can further improve the robustness to noise on the basis of fast and stable reconstruction.In silico model experiments verify that the reconstruction algorithm has better noise robustness and higher reconstruction accuracy.（3）In order to further improve the speed of MFMT reconstruction,a sparse reconstruc-tion algorithm based on adaptive fast iterative shrinkage-thresholding is proposed.Taking into account the sparseness and non-negativity of the distribution of specific fluorescent markers in the tissue,the regularization term of the mixture of L₁and L_1/2is introduced as the penalty term of the fast iterative shrinkage-thresholding algorithm（FISTA）,and adopt an adaptive method is used to select the regularization factor to achieve faster and more ac-curate sparse reconstruction.In silico model experiments and vessel tree model experiments have proved the effectiveness of the proposed algorithm in the reconstruction of fluorescent probes spatial distribution at the mesoscopic scale.It can not only obtain high-precision reconstruction results while further improving the reconstruction speed,but also has good robustness to noise.