Abstract

ABSTRACT

Historically, dosimetry has been conducted with varying degrees of sophistication. A broad and general approach involves computing organ-specific time-integrated activity multiplied by S-factors, as defined by the Medical Internal Radiation Dose (MIRD) Committee. This method uses a standard human body phantom as a reference for mass density distribution. More advanced methods, such as the voxel dose approach, estimate absorbed dose by linearly superimposing contributions from each voxel in the spatial activity distribution, treating each voxel as a radiation point source. The energy dose from a radiation point source with isotropic unit activity in an infinite homogeneous medium is known as a dose point kernel (DPK). Previously, DPKs were calculated using Monte Carlo techniques for tissues like bone, lung, soft tissue, and water, primarily for isotopes such as Iodine-131 and Yttrium-90. Later, continuous DPKs were discretized into dose voxel kernels (DVKs) that can be arbitrarily scaled. The absorbed dose distribution is then derived by convolving the time-integrated spatial activity distribution with the Monte Carlo-based DVK, based on the patient's anatomy. Due to the computational expense of this process, various methods have been explored to estimate approximate DVKs. The highest level of sophistication is achieved with full Monte Carlo simulations of radiation transport inside the patient's body, providing reference results for benchmarking other approaches. The network is trained with DVKs obtained from dedicated Monte Carlo simulations using equally sized kernels of specified tissue density and a specified radioisotope. This method is intermediate between the canonical MIRD protocol and deep neural network approaches that predict whole-body absorbed dose distributions from individual mass density and activity distributions. In this paper, we have discussed and compared all the commonly employed dosimetric method along with new approaches using artificial intelligence.

Key words: dose voxel kernels, organ-specific time-integrated activity, Medical Internal Radiation Dose.