Validating a grid-based Boltzmann solver for nuclear medicine applications in homogeneous and heterogeneous materials for voxel-based absorbed doses

Objectives In radiation oncology, deterministic methods of radiation transport have been translated into clinical practice. Here we investigated a deterministic grid-based Boltzmann solver (GBBS) for NM absorbed dose calculations in homogeneous and heterogeneous media.

Methods Voxel-S-values were calculated for monoenergetic β and γ (1, 0.1, 0.01 MeV), 90Y, and 131I for 3 mm voxels using Monte Carlo (MC, DOSXYZnrc) and GBBS (Attila v8.0.0) with infinite slabs of soft-tissue (S, ρ =1.04 g/cc), lung (L, ρ=0.26 g/cc), and bone (B, ρ = 1.85 g/cc). Heterogeneity was evaluated by performing simulations with a single source voxel at an interface; we investigated S→L, L→S, S→B, and B→S where the first material represented the source voxel material. We report the percent difference (%Δ) of GBBS from MC (gold-standard) in the source voxel for uniform media, and the source voxel and adjacent voxel across interface for the heterogeneous media. Radial plots of voxel-S-values comparing GBBS with MC were made for the uniform media. For heterogeneity simulations, radial plots of voxel-S-values for each plane on either side of the interface were generated to compare GBBS and MC.

Results %Δ in S, L, and B absorbed doses between GBBS and MC ranged from -6.4 to -3.3 for 1 MeV β, -8.1 to -7.5 for 0.1 MeV β, -13.6 to -13.4 for 0.01 MeV β, -5.5 to -4.5 for 90Y β, -9.7 to -6.6 for 131I β, -7.1 to -6.6 for 1MeV γ, -9.8 to -9.1 for 0.1MeV γ, -19.6 to -15.2 for 0.01MeV γ, -13.6 to -7.3 for 131I γ. For the interface simulations the %Δ in the source voxel and adjacent voxel in other material were similar in magnitude to those observed in uniform media. Radial plots for both the uniform and interface planes showed that GBBS and MC were in good agreement.

Conclusions GBBS can generate absorbed dose maps in both homogeneous and heterogeneous media with clinically acceptable uncertainties. GBBS has the potential for faster computation and radionuclide-specific optimizations.

Research Support NIH/NCI R01CA138986