Research On Preprocessing Strategies For Bioluminescence Tomography

Bioluminescence tomography(BLT)is an important imaging method of molecular imaging.It uses CCD camera to collect photon energy density data on the surface of organisms and combines with the propagation model of light in organisms to reconstruct the distribution of light source in organisms.As a highly sensitive non-invasive imaging technique,it can be developed to detect and diagnose diseases in their early development.At the same time,it can also be used in drug research,therapeutic efficacy evaluation and other fields to reduce the cost of drug research and shorten the development cycle.It has a wide range of medical applications in biomedicine and pre-clinical research.At present,the research on bioluminescent tomography mainly focuses on the establishment of transmission model,reconstruction algorithm and application.However,due to the serious ill-posedness of the BLT inverse problem,the reconstruction process is very sensitive to model errors,discrete errors and unavoidable noises.To reduce the ill-posedness of the reconstruction problem and improve the stability and accuracy of the reconstruction,this paper preprocesses the system matrix based on the finite element analysis and the multi-spectral measurement acquired by the imaging system respectively from both sides of the BLT system equation.Two kinds of pretreatment strategy are proposed:(1)By singular value decomposition of the coefficient matrix A obtained after discretization,it can be found that the singular value of A tends to zero gradually,and the condition number of A is large.Therefore,according to the matrix decomposition theory,we adopt the truncated singular value decomposition preconditioning strategy.A new coefficient matrix is constructed by removing the small singular value in the coefficient matrix which is easy to cause obvious errors.And on this basis,the light source is reconstructed by combining with the sparse reconstruction algorithm.(2)Considering the systematic errors and measurement noises caused by the different positions of organisms and the characteristics of the imaging system during the experimental operation,the second preprocessing strategy--logarithmic ratio of light intensity method is adopted in this paper.In order to reduce the ill-condition of the reconstruction and the experimental error,the logarithm and ratio of the obtained data of multiple near-infrared spectral segments are calculated.It is proved that the preprocessing strategy can effectively reduce the ill-condition of the inverse problem by the simulation results.