Bioluminescence　tomography　(BLT)　is　one　of　the　most　important　non-invasive　optical　molecular　imaging　modalities.　The　model　for　the　bioluminescent　photon　propagation　plays　a　significant　role　in　the　bioluminescence　tomography　study.　Due　to　the　high　computational　efficiency,　diffusion　approximation　(DA)　is　generally　applied　in　the　bioluminescence　tomography.　But　the　diffusion　equation　is　valid　only　in　highly　scattering　and　weakly　absorbing　regions　and　fails　in　non-scattering　or　low-scattering　tissues,　such　as　a　cyst　in　the　breast,　the　cerebrospinal　fluid　(CSF)　layer　of　the　brain　and　synovial　fluid　layer　in　the　joints.　A　hybrid　Radiosity-diffusion　model　is　proposed　for　dealing　with　the　non-scattering　regions　within　diffusing　domains　in　this　paper.　This　hybrid　method　incorporates　a　priori　information　of　the　geometry　of　non-scattering　regions,　which　can　be　acquired　by　magnetic　resonance　imaging　(MRI)　or　x-ray　computed　tomography　(CT).　Then　the　model　is　implemented　using　a　finite　element　method　(FEM)　to　ensure　the　high　computational　efficiency.　Finally,　we　demonstrate　that　the　method　is　comparable　with　Mont　Carlo　(MC)　method　which　is　regarded　as　a　＇gold　standard＇　for　photon　transportation　simulation.