Rapid internal 3-dimensional dose calculation for 2,174 radionuclides based on ICRP-110 reference phantoms

Introduction: Standard internal dosimetry with high accuracy is crucial for determining first-in-human dose for clinical trial. The direct Monte Carlo (MC) simulation is regarded as a ground-truth voxel-based dosimetry technique, but it has a high computational cost. To overcome the limitations of the direct MC approach, we propose a fast internal dose calculation platform for 2,174 radionuclides (RI), which applies a dose point kernel to estimate the whole-body voxel dose on ICRP-110 reference phantoms.

Methods: We used dose point kernels (DPK) for 2,174 RI and the ICRP phantoms. Nuclear data of 2,174 RI were acquired from the NNDC. Graves et al. calculated the DPK from isotropic point sources of activity in water using MC simulation. The data of RI included the emission type for photons (γ-rays, x-rays), beta particles (positrons, electrons); and discrete electrons (conversion electrons, Auger electrons, Coster–Kronig electrons). Alpha particle, neutron, and spallation effects were not considered. The adult male phantom is composed of 254 × 127 × 222 voxels of size 2.137 x 2.137 x 8.0 mm. The adult female phantom is composed of 299 x 137 x 348 voxels of size 1.775 x 1.775 x 4.84 mm. The voxel phantoms have organs/tissues that can be defined as the source and target region in the gastrointestinal tract, urinary bladder, the surfaces of tissues in the skeleton, the alimentary tract, and the respiratory tract. Our platform provides the options for users to select the type of voxel phantoms (male vs female), source organ, target organ, and RI (number and emission type). The RI are randomly selected in the x, y, and z-axis indices of the target organ and are located at the center of the selected voxels. We created the 3-dimensional (3D) dose map using the dose per distance between the centers of voxels. Our platform can also provide a dose volume histogram for the target organ. We added the option for users to define the voxel resolution to increase the accuracy of the calculation.

Results: Our platform has the function to provide the visualized 3D dose map with voxel phantoms. Users can confirm the dose map information per slice by scrolling for the z-axis direction (from head to toe). We calculate the dose map with setting option (phantom: male, radionuclide: Lu-177, source organ: liver, target organ: liver, the number of RI: 1,000) (Fig. 1. (a)) and the cumulative dose volume histogram for left lung tissue (Fig. 1. (b)). We validated our platform with OpenDose which provides S-values based on MC simulation and ICRP phantoms. After calculating S-values using Lu-177 for the beta emitter and Tc-99m for the gamma emitter, they were compared with S-values of OpenDose (Fig. 2). The mean difference with OpenDose is about 10% and the mean calculation time for target organs is about 6 minutes. Relatively large differences between our platform and OpenDose were observed when the source organ and target organ are the same, and the target organ is the wall organ due to the partial volume effect. By increasing the voxel resolution, we quantified the errors of the self S-values of the liver and gall bladder wall. We also calculated the computation time per voxel resolution for both cases (Fig. 3). The mean errors of the S-values for the liver and gall bladder wall were reduced by approximately 40% and 36%, respectively when the voxel resolution was improved from 27 / cm3 to 739 /cm3. The mean computation time of both cases increased from approximately 6 minutes to 3 hours. They indicate that the higher voxel resolution reduces the error of the S-value, but it increases the computation time.

Conclusions: We developed the rapid internal dosimetry platform that provides a 3D dose map based on ICRP phantoms and 2,174 dose point kernels. Our platform can offer users the opportunity to calculate the S-values for the clinical trial of radiopharmaceutical therapy and emergency exposure situations without time-intensive MC simulation. Our platform will be distributed as open-source code.

• Download figure
• Open in new tab
• Download powerpoint