VoxIsle: A new method in InterDosi code for improving internal dosimetry simulation performances in voxelized phantom

Abstract

In order to assess the internal dose due to exposure to radiation, the ICRP recommends the use of a reference voxelized phantom. Although voxelized geometry is a very promising geometric representation of a complex body, it showed a limitation in dose calculation for large-voxel array phantoms that requires more memory and CPU time. We have developed a new voxel compression method called VoxIsle that merges adjacent voxels that belong to the same organ/tissue. The VoxIsle was based on the LRAH (Largest Rectangular Area Histogram) algorithm. To evaluate the performances of this method, compressed and uncompressed geometries of some voxel-based phantoms were used to calculate dosimetric coefficients of gamma, electron and alpha particles. Results showed that for a given particle, there is a smooth difference between coefficients calculated using the two types of geometry. Additionally, depending on the characteristics of the voxelized phantom, the memory and CPU time required to construct geometry and material of a phantom, and its file size for a compressed geometry appear to be reduced up to 77%, 73% and 97%, respectively. Regarding computational performance, with the exception of alpha particles for which the results showed a steady increase in CPU time, the simulation with compressed geometry was up to 59.3% faster than the uncompressed geometry. The new compression method can considerably improve the simulation performances of InterDosi code that can perform internal dosimetry studies on voxelized phantoms.