Localizing source distribution based on the adaptive finite element methods in bioluminescence tomography

In small-animal molecular imaging, bioluminescence tomography (BLT) is used to reconstruct the internal bioluminescent source which can reveal the molecular and cellular information. Based on the general finite element tomographic method, the finer the discretized mesh to the given domain is, the more precise the BLT solution may be. However, the expensive computational cost will be utilized for BLT reconstruction. In this contribution, an adaptive finite element methods based tomographic algorithm is represented and used to localize bioluminescent source. In the proposed algorithm, an initial coarse volumetric finite element mesh is provided and a priori knowledge is employed to determine the permissible source region. Furthermore, the local mesh refinement is performed to adaptively reduce the element size of the mesh by virtue of error estimation techniques. The above strategies reduce the ill-posedness of the BLT problem significantly and improve the numerical stability effectively. Numerical simulations with the homogeneous and heterogeneous phantoms, in which the synthetic data is obtained through the adaptive finite element solver and Monte Carlo methods, show the effectiveness of the tomographic algorithm.