Adaptive angular resolution for time-dependent radiative transfer simulations based on local radiation field anisotropy

Developing a fast and accurate radiative transfer calculation method is vital in modeling light propagation in biological tissues. We present a novel acceleration scheme for three-dimensional time-dependent radiative transfer simulation. In biological tissues, required angular resolutions for anisotropic radiation fields sensitively depend on spatial positions. Our new scheme changes the angular resolutions with the positions automatically based on local anisotropies of the radiation fields using the spherical Haar wavelet basis function. We apply the new code to simulations of light propagation in a box of polyurethane phantom mimicking biological tissue. We show that the new scheme achieves the acceleration of a factor of ∼ 8 at most and ∼ 3 on average compared to a simulation with the constant angle resolution while its accuracy keeps the same level. We find that the acceleration becomes more noticeable if low-angular resolution areas spread with time according as the diffuse radiation dominates. As the result of the fewer required angle bins, we find that the new code has the potential to reduce the memory use to ∼ 20 % at the maximum. Thus, our new scheme also has the advantage from the viewpoint of computational resources. We note that our new scheme is applicable not only to steady-state media but also to scenarios where the physical state of matter changes dynamically. Our new scheme will be a powerful tool to perform radiative transfer simulations of more than a thousand models which must be important in developing machine learning models.