Voxelized scomber japonicus phantom and its application for internal dose evaluation by Monte Carlo calculations

This study developed a voxelized scomber japonicus phantom through CT scanning images for the purpose of radiation protection in the environment, which gives a realistic description of the scomber japonicus’s anatomy. The absorbed fraction (AF) values of the voxelized scomber japonicus phantom were calculated for photons and electrons in the energies of 0.01, 0.02, 0.03, 0.04, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 1.0, 2.0, 3.0, 4.0 and 5.0 MeV using Monte Carlo toolkit Geant4. Furthermore, the S-factor values for four radionuclides (134Cs, 137Cs, 131I and 90Sr) in the organs/tissues were evaluated using the monoenergetic AF values based on the radioisotope decay data, and compared with that obtained directly through whole emission spectra simulation. The voxelized scomber japonicus phantom would be a good candidate for dosimetry in non-human biota for the environmental dosimetry, and the built database of AF values can be used for dose calculations of any radioactive contaminations.


Introduction
In order to gain a comprehensive understanding of the dose-effect relationships in non-human biota, it is crucial to conduct precise assessments of organ doses in animals, which can provide the fundamental basis for establishing appropriate protection criteria. The absorbed fraction (AF) is commonly utilized as a dose conversion factor in the assessment of internal radiation, representing the proportion of energy emitted by a radiation source that is absorbed within the target organ or tissue. In addition, estimation of organ doses may be accomplished through the utilization of S-factor values (Gy/Bq/s), as defined by the Medical Internal Radiation Dose (MIRD) committee. These values represent the mean absorbed dose to a target organ per unit of cumulative activity in the source organ, and can be derived from the AF value. The employment of voxelized phantoms in conjunction with Monte Carlo simulation is a viable approach to computing the AF and S-factor values. International Commission on Radiological Protection (ICRP) has proposed 12 reference animals and plants [1] for the purpose of evaluating the impact of radiation on non-human biota, including earthworm, duck, flatfish, crab, brown seaweed, trout, frog, deer, pine tree, grass, bee, and rat. Additionally, ICRP has assessed the AF values for non-human biota [2].
This study developed a voxelized scomber japonicus phantom as a non-human model for radiation protection in the environment. The Monte Carlo toolkit Geant4 was utilized to conduct radiation transport calculations of AF values with the voxelized scomber japonicus phantom for monoenergetic photons and electrons, resulting in a database that can be utilized for dose calculations of any radioactive contaminations. Moreover, considering the enormous emissions of artificial radionuclides ( 134 Cs, 137 Cs, 131 I and 90 Sr) after the Fukushima Daiichi Nuclear Power Plant accident [26,27], the S-factor values for these radionuclides were calculated using AF values, and compared with that directly calculated through simulation of the whole emission spectra.

The voxelized scomber japonicus phantom
The scomber japonicus used in this study was physically captured from the sea near the Tianwan Plants in Jiangsu Province, China. It weighted approximately 243 g and had a total length of 26.1 cm, as shown in figure 1(a). The scomber japonicus was immediately frozen after capture and imaged in a frozen state.
To create the voxelized scomber japonicus phantom, four steps were utilized as illustrated in figure 2. Firstly, a set of medical images was acquired using computed tomography (CT) or magnetic resonance imaging (MRI). In this study, the CT scanning was employed to obtain the anatomic data in DICOM files, which was performed at Jiangsu Cancer Hospital on a Siemens SOMATOM Definition AS 64 slice machine, with helical acquisition of images at 120 kVp and 50 mA, and a slice thickness of 1 mm. Figure 1(b) presents a slice image of the CT scan. Secondly, contour information for each organ/tissue of the scomber japonicus was segmented from the CT images, and the voxelized phantom was created using 3D Slicer software [28]. The segmentation process involved contouring individual organs/tissues on each slice of the CT image based on their greyscale range, and generating 3D models by stacking the 2D contours, which were then exported as OBJ files for further processing. Thirdly, the 3D models in OBJ format were voxelized using the open-source program Binvox [29], which generated binary voxel (BINVOX) grid files. The consistency of the BINVOX grid file was checked using the open-source program Viewvox [30], and the BINVOX grid file was further converted to ASCII TXT file. Lastly, an in-house code [31] was developed to integrate each organ into the final whole voxelized phantom, with each organ/tissue assigned an index. Table 1 presents a comprehensive list of the densities and elemental compositions of the 14 organs/tissues represented by index 1 to index 14 in the voxelized scomber japonicus phantom. These indexes correspond to the following organs/tissues: skin, bone, brain, eyes, gall bladder, gill, gonad, heart, kidney, liver, rectum, stomach, swim bladder, and muscle. The information for the swim bladder is sourced from Saunders's work [32] as same with Ruedig [21], whereas the data for the remaining organs/tissues are obtained from ICRU report 44 [33], 46 [34] and ICRP publication 89 [35].

Monte Carlo simulation
Monte Carlo simulation toolkit Geant4 [36] was used in this research, which is widely used in radiation physics research spanning high energy physics, medical physics, and space science, due to its exceptional flexibility, capability of multiple particle transport, and open-source nature.
For modelling and calculations, three physics-list libraries were utilized, namely G4EMStandardPhysics, G4DecayPhysics, and G4RadioactiveDecayPhysics. Specifically, G4EMStandardPhysics was used to simulate the electromagnetic process, including photoelectric effect, Compton scattering, pair production, multiple scattering, ionization, and bremsstrahlung for photons and electrons; G4DecayPhysics was used for modelling decays of unstable particles defined in the physics list; G4RadioactiveDecayPhysics was used to simulate radioactive decays of nuclei based on ENSDF data [37]. To speed up the simulation and reduce memory usage, the G4PhantomParameterisation class was used to create the parameterized volume representing the voxelized scomber japonicus phantom, and deposited energy was scored by organ/tissue index in each particle tracking step instead of scoring the three-dimensional dose distribution [38]. The Geant4 toolkit version 11.0.3 was used in the present work.
To simulate the internal exposures, the organs/tissues inside the phantom were assumed as the source of ionizing radiation. And the source was respectively set as uniformly distributed monoenergetic photons and electrons with 0.01, 0.02, 0.03, 0.04, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 1.0, 2.0, 3.0, 4.0, and 5.0 MeV. The particles deposited all their energy in the surrounding tissues, directly irradiating internal organs. The energies deposited in the different organs/tissues were obtained through the Monte Carlo simulation. The number of primary particles was set as 1 × 10 8 for the simulations in this study to reduce the uncertainty in the data to be less than 5% for most cases, and the data whose uncertainty exceeded 20% are not considered. The uncertainty is calculated according to Seco et al [39] and Chauvin et al [40]. All the simulations were performed on a computer with 64 cores (Intel Xeon 8375 C CPU @ 2.90 GHz × 64) running Ubuntu 20.04.3, and each simulation required a CPU time of about 0.5 h.

Calculation of AF, self-AF and S-factor values
Based on the energies deposited in the target organs from Geant4 simulations, the AF value can be calculated as follows, is the energy deposited in the target organ t obtained from Geant4 simulations in MeV, E . o is the initial energy released by all particles in the source organ (MeV), and AF is the absorbed fraction of the energy in the target organ t derived from radiation emitted from source organ s. As for the self-AF value, it specifically refers to the AF values when the source organ is the same as the target organ.
For a given radionuclide, the S-factor value can be calculated as follows, where i is the radioactive particles of different energies emitted from the given radionuclide, y i and E i are respectively the branching fractions and energies of the particles, ( ( )) ¬ E t s i d is the absorbed energy, either  simulated or interpolated, resulting from particles emitted with energy E , i and m t is the mass of target organ t in kg. The factor 1.6 × 10 −13 is used to convert unit of the S-factor values to Gy/Bq/s. S-factor values denote the mean absorbed dose that a target organ receives per unit activity of a radionuclide present in the source organ. These values are unique to each radionuclide and are utilized to estimate radiation doses in various organs and tissues. Monte Carlo methods are widely employed for the estimation of S-factor values by simulating radiation transport in phantoms. Two methods are available for estimating S-factor values specific to radionuclides. The first method, referred to as the direct method, entails the direct evaluation of emission spectra through Monte Carlo simulations. The second method, known as the integration method, involves the summation of contributions from each branch of photons and electrons. In the integration method, the AFs for individual branching energies of the radionuclides are acquired through interpolation from precalculated monoenergetic AF values. These interpolated AFs are then multiplied by their corresponding branching fractions and subsequently summed to derive the S-factor values. In order to assess the efficacy of the AF values database for computing doses resulting from any radioactive contaminations, a comparative analysis was conducted for 134 Cs, 137 Cs, 90 Sr, and 131 I, utilizing S-factor values obtained through direct method or integration of monoenergetic results across the whole emission spectra. The photon and electron emissions data utilized in this study for the selected nuclides were sourced from ICRP Publication 107 [41]. The particles listed in table S1 were considered for the respective nuclides of 134 Cs, 137 Cs, 90 Sr, and 131 I in this study. Figure 1(c) presents a detailed representation of the voxelized scomber japonicus phantom that is created using 3D Slicer software. The phantom is constructed with a voxel size of 0.5 mm, a matrix consisting of 552 ×116 × 77 elements, and 14 segmented organs/tissues, including skin, bone, brain, eyes, gall bladder, gill, gonad, heart, kidney, liver, rectum, stomach, swim bladder and muscle. However, as the skin and muscle of the phantom tend to suppress the internal structures, they are not visible in figure 1(c) to highlight the individual segments. Considering the difficulty to distinguish the boundaries between some organs, in this study, skin and scales are considered as organ skin, moreover, fins, pharynx, thyroid, spinal cord, nerves, vasculature, muscle, and miscellaneous soft tissue are all categorized as tissue muscle.

Voxelized scomber japonicus phantom
The volume of each organ/tissue can be determined by calculating the number of voxels of each organ/ tissue and the volume of single voxel after the voxelization and integration process. The mass of each organ/ tissue is then calculated by multiplying the volume with the density of each organ/tissue. Table 2 presents the corresponding voxel numbers, volumes, and masses of each segmented organ/tissue for the voxelized scomber japonicus phantom. The phantom's total mass is calculated to be 245.116 g, which is approximately 0.871% greater than the actual mass of the scomber japonicus (243 g). This discrepancy is deemed acceptable given that the organ/tissue compositions and densities employed in this study were based on human rather than fish organ/tissue.  Figures 3 and 4 present the self-AF values for photons and electrons, respectively, in the voxelized scomber japonicus phantom over a range of energies from 10 keV to 5 MeV. Here, the source and target organs in both cases are identical. Figure 3 illustrates a general trend of decreasing self-AF values as photon energy increases. Specifically, the self-AF values for photons demonstrate a sharp decline from 10 keV to 100 keV, indicating a rapid increase in photon escape from the source organs as energy increases within this range, with the local minimum at 100 keV (except for bone and swim bladder, which presents a local minimum at 200 keV). Following these relative minimum values, the self-AF values experience a slight increase in the energy range of 100 keV (200 keV for bone and swim bladder) up to approximately 1 MeV, followed by a decline from 1 MeV to 5 MeV. On the other hand, figure 4 illustrates that the self-AF values for electrons remain stable up to several tens  of keV before decreasing monotonically with increasing energy. Nevertheless, as the initial electron energy increases from several tens of keV to 5 MeV, the self-AF values decrease significantly. The swim bladder behaves differently than other organs in terms of self-AF values, likely due to its unique structure. The swim bladder contains gaseous contents, making it less dense than other organs in the scomber japonicus. This lower density results in a reduced self-AF value, since there is less likelihood of photon and electron interactions in gaseous media.  Figure 7 represents the case where the source organ is the liver, while figure 8 represents the bone as the source organ. Table S3 in the Supplementary Materials provides a more comprehensive view of the AF values and uncertainty of electron sources for the various tissues/organs.

Discussion
Considering the radiation protection requirements in the vicinity of the Tianwan Plant in China, a voxelized scomber japonicus phantom was developed in this study to establish a comprehensive database for calculating dose from radioactive contamination.
The self-AF values for photons and electrons are presented in figure 3 and figure 4, respectively. The trends observed in figure 3 are consistent with the properties of photon interactions. Specifically, the photoelectric effect is the dominant process at low photon energies (tens of electron volts up to a few hundreds keV), leading to a majority of the photon's energy being retained in the source organs. This phenomenon causes a local   figure 3. With further increase in photon energy, pair production becomes the dominant process, leading to a decline in self-AF values from 1 MeV to 5 MeV, as pair production gammas escape the target organs. Meanwhile, the trend depicted in figure 4 also meets with the properties of the electron interactions. Specifically, at electron energies below approximately 100 keV, the source organs exhibit nearcomplete absorption of the emitted electrons. It shows that the particle escape from the source organs initially grows slowly with energy compared to the photon self-AF values in figure 3. As the electron energy increases, the self-AF values experience a significant decrease. It is worth noting that, in the case of small non-human biota, not all released electron energies are deposited in the source organ, which is similar to the situations of small organs in human dosimetry.
Overall, the self-AF values for both photons and electrons are dependent on the energy of the radiation. Disparities in self-AF values at identical energy levels are predominantly affected by variations in the mass of the source organs. Specifically, a greater mass of the source organ is positively correlated with a higher self-AF value at a given energy. For instance, at the same energy level, the self-AF value for muscle (153.644 g) exceeds that of bone (46.397 g). Figure 10 presents a comparative analysis of the self-AF values for the liver as a source organ between the previously established rainbow trout phantom [21] and the scomber japonicus phantom developed in this study. The results reveal a similar trend in both photon and electron sources, albeit with observed differences in self-AF values that may be attributed to variations in phantom mass. Specifically, the mass of the rainbow trout phantom, with a mass of 658 g, is considerably heavier than the scomber japonicus phantom. This finding underscores the importance of developing phantoms for various types and sizes of non-human biota. The biggest difference between the common shape of curves of the self-AF values in figure 3 and the AF values in figure 5 (excluding the liver, which serves as source organ) occurs in the low energy ranges. Specifically, a discernible upsurge is observable in the low energy region of figure 5, which is diametrically opposed to the pattern observed in figure 3. This phenomenon is attributed to the escalation in the number of particles that escape from the source organ, resulting in an increase in the number of particles that deposit energy in the several adjacent target regions as the initial photon energy rises. The initial peaks observed in figure 5 for each target organ can be ascribed to a larger number of particles depositing energy outside the organ of interested or escaping from the phantom beyond a certain initial photon energy.
It can be noticed from figure 5 that AF value for the liver is nearly 1 at the lowest energy (0.01 MeV), indicating that almost all the absorbed energy remains in the source organ, i.e., liver. As the photon energy increases, a greater amount of energy is deposited in organs other than the liver. Specifically, the majority of the deposited energy shifts from the liver to the muscle as the photon energy increases to 0.03 MeV. In most target organs, the AF values initially increase rapidly with energy, and then slightly decrease with further increases in energy. Moreover, photons with energies above 0.1 MeV can travel greater distances compared to the size of the organs.
The AF values of photons in figure 5 and figure 6 decrease monotonically with energy, while the AF values of electrons in figure 7 (except for liver, which is source organ) and figure 8 (except for bone, which is source organ) increase monotonically with energy. In principle, regardless of the energy of the electron source, nearly 100% of the energy will be deposited in the fish body. It can be noticed that when the electron energy is 0.01 MeV, almost 100% of the absorbed energy is in the source organ. However, as the electron energy increases, the AF values in source organs decrease gradually while those in target organs increase gradually. The position of the target organ relative to the source organ is the primary indicator of AF values in the target organ due to electron interactions. Besides the source organ, those organs positioned closest to the source have the largest AF value.
Based on these results from monoenergetic simulations, the S-factor values for radionuclides can be obtained. The accuracy of the database created is assessed by comparing the S-factor values for 134 Cs, 137 Cs, 131 I, and 90 Sr obtained by direct calculation and integration, as shown in figure 10. The results reveal a good agreement between the two methods. The differences between the methods can be explained partly by the uncertainty accompanying each value from Geant4 simulation, the error from the interpolations, and the difference of emission spectra (including branching fractions and energies) of radionuclides between direct Geant4 simulation and integration. The S-factor values for self-irradiation also depend on the mass of the source organs. Consequently, the developed database predicated on monoenergetic photons and electrons may be employed for expeditious evaluation of environmental dosimetry in the vicinity of the Tianwan Plant in China.
There are limitations to the current study. While the scomber japonicus is extensively distributed in the northwestern, southeastern, and northeastern Pacific, the database generated in this study may not be representative of other marine organisms or regions. To study radiation doses in more species and regions, it is necessary to investigate the situation in a wider variety of marine organisms. Furthermore, disparities in the elemental compositions and densities of scomber japonicus tissues and those of humans may impact the accuracy of our dosimetry calculations. Given the potential benefits of MRI scans in yielding greater detail of soft tissues compared to CT scans, the integration of CT-MRI fusion could be a promising approach in furnishing more precise and comprehensive insights into the structures and functions of animals in subsequent research. Additionally, the ongoing development of a mesh-type phantom is expected to depict organ shapes more accurately, particularly for thin and small organs, relative to voxelized phantoms. Furthermore, experimental validations could also be performed in the future.

Conclusions
A voxelized phantom of the scomber japonicus was developed using CT images of scomber japonicus which physically caught at the sea area near Tianwan Plants in Jiangsu Province, China. The phantom accurately represented the animal's anatomy with similar masses and sizes as the original. The self-AF and AF values for monoenergetic photons and electrons in the organs/tissues of the voxelized scomber japonicus phantom were evaluated with Monte Carlo toolkit Geant4 in the energy range from 0.01 to 5 MeV. Furthermore, the obtained AF database was used to evaluate the S-factor values for 134 Cs, 137 Cs, 131 I and 90 Sr in the organs/tissues of the phantom. And these S-factor values were comparable to those obtained through direct Geant4 simulation. The voxelized scomber japonicus phantom has potential for use in evaluating internal organ dose in non-human biota exposed to ionizing radiation.