Study On Sparse Constraint Regularization Of Bioluminescence Tomography

In recent years，molecular imaging methods have been broadly applied inmedical diagnosis because it can monitor nondestructively and dynamically. BLTwhich is a rising molecular technology can monitor physiological changes of thetissues in molecular level， and are mainly used to observe tumor cells growth andmetastasis. In view of this feature，bioluminescence tomography will be a quitepowerful tool for malignant disease prevention and diagnosis.The Radiative Transfer equation (RTE) can depict photon propagation processflexibly. However it is difficult to solve， most of the research are based on thediffusion approximation equation (DA) of RTE. To light source inversion， DAcan’t produce ideal results in the region that near the light source or relatively smallscattering coefficient.In BLT， one tries to recover the distribution of isotropic bioluminescentsources based on optical information from boundary measurement. Theoreticalsystem of BLT has been established completely after years of research. But becauseof insufficient measurement and other facts， the solving process of bioluminescentsources is ill-posed， the reconstruction algorithm must be able to overcomeill-posedness， have ideal computational efficiency and good reconstruction results.Angle and space discretization lead to long time calculation. To thecharacteristics of ill-posed， large calculating and sparse distribution of light sourceof BLT， we choose l1-regularization with Split-Bregman method. In simulation，we use l1-regularization and l2-regularization to invert internal sourcerespectively， judging from the results， l1-regularization performs in generalbetter in source inversion with only boundary data.