Direct Monte Carlo Dose Calculation Using Polygon-surface Computational Human Model

Abstract

Two different kinds of computational human models are available for Monte Carlo dose calculations - stylized models based on mathematical surfaces and voxel models based on tomographic images. Recently, some investigators developed hybrid-type models by converting their voxel models to polygon or non-uniform rational B-spline (NURBS) surface models. These surface models have both the flexibility of the stylized models and the realism of the voxel models. The surface model, however, has a critical limitation; that is, a surface model should be converted back to a voxel model to be used for Monte Carlo dose calculations. This so-called "voxelization" process eliminates several advantages of the surface model, including the capability of modeling very thin structures and the capability of 4-D Monte Carlo simulation. In the present study, a polygon-surface human model was directly imported to a Monte Carlo code, Geant4, and used for organ dose calculations. For this, the G4TessellatedSolid class was used, which was originally developed to import computer-aided design (CAD) models. The original voxel model was also imported to Geant4 to compare computational speed and calculated dose values. The average size of the polygons of the polygon-surface model was ~0.5 mm2 while the voxel resolution of the voxel model was 1.981 × 1.981 × 2.0854 mm3. The calculated dose values from the polygon-surface and voxel models were in a good agreement; that is, the difference was insignificant for the photon energies and irradiation geometries considered in the present study. The polygon-surface model was slower than the voxel model by factor of ~10 times in dose calculation, which is acceptable considering that the direct use of the polygon-surface model does not use a separate voxelization process which also takes some time.