| Bioluminescence Tomography is one of the method in optical molecular imaging, via the construction of the source, the cancer can be found in cellular levels in advance. Due to the lock of measurement on the surface and the complexity of the photon transfer in the tissue, the source reconstruction was turned to an ill-posed problem. How to reconstruct the source distribution was the core problem of BLT.The paper which focuses on the source reconstruction of the BLT, was intended to get a robust result through a suitable algorithm. We armed at the ill-posed character and reconstructed the source by two kinds of method which was the greedy method based on Compressed Sensing (CS), and some regularization algorithms based on the ι1norm. Some work was done as below:1) From the perspective of digital signal processing and inspired by the CS, we put the source reconstruction as a problem of sparse signals recovery. We reconstructed the sources through4greedy algorithms. They were Orthogonal Matching Pursuit (OMP), Stage-wise OMP (StOMP), Regularized OMP (ROMP) and Compressive Sampling MP (CoSaMP). We verified the effectiveness of the algorithms for the problem by simulation experiments, and assess their performance.2) We convert the problem to be an equation of underdetermined linear system and optimized it through4ι1norm algorithms. We subdivided the finite element grid combined with the self-adaption theory and shrieked the scope of the permissible source region, then the source would be reconstructed on each level of grid. We could get some index including not only the accurate location but also the energy density and total energy for assessment. To verify the effectiveness of the algorithms we have done some experiments in a3D digital mouse model. As the experiments showing, each of the algorithms could reconstruct the source accurately with some difference in the energy density and accuracy of source location. We analyzed the difference and gave some advice to the further research. |
PDF Full Text Request