At present, the most widely accepted forward model in diffuse optical tomography (DOT) is the diffusion equation,
which is derived from the radiative transfer equation by employing the P1 approximation. However, due to its validity
restricted to highly scattering regions, this model has several limitations for the whole-body imaging of small-animals,
where some cavity and low scattering areas exist. To overcome the difficulty, we presented a Graphic-Processing-
Unit(GPU) implementation of Monte-Carlo (MC) modeling for photon migration in arbitrarily heterogeneous turbid
medium, and, based on this GPU-accelerated MC forward calculation, developed a fast, universal DOT image
reconstruction algorithm. We experimentally validated the proposed method using a continuous-wave DOT system in the
photon-counting mode and a cylindrical phantom with a cavity inclusion.