We present an algorithm using data acquired with a time-resolved system with the goal of reconstructing sources of fluorescence emanating from the deep interior of highly scattering biological tissues. A novelty in our tomography algorithm is the integration of a light transport model adapted to rodent geometries. For small volumes, our analysis suggest that neglecting the index of refraction mismatch between diffusive and non-diffusive regions, as well as the curved nature of the boundary, can have a profound impact on fluorescent images and spectroscopic applications relying on diffusion curve fitting. Moreover, we introduce a new least-squares solver with bound constraints adapted for optical problems where a physical non-negative constraint can be imposed. Finally, we find that maximizing the time-related information content of the data in the reconstruction process significantly enhances the quality of fluorescence images. Preliminary noise propagation and detector placement optimization analysis are also presented.