Improved Accuracy of S-value based dosimetry: Transitioning from Cristy-Eckerman to 1 ICRP adult phantoms 2

In 2016, the International Commission on Radiological Protection and Measurements (ICRP), published 3 the results of Monte Carlo simulations performed using updated and anatomically realistic voxelized 4 phantoms. The resulting absorbed fractions are substantially more accurate than calculations based on the 5 Cristy-Eckerman (CE) stylized (or mathematical) phantoms. Despite this development, the ICRP absorbed 6 fractions have not been widely adopted for radiopharmaceutical dosimetry. To help make the transition, we 7 have established a correspondence between tissues defined in the CE phantom and those defined in the 8 ICRP phantoms. Using pre-clinical data from biodistribution studies performed, we have calculated 9 absorbed doses for Th-227 labeled HER2 targeted antibody. We compare the CE phantom-based 10 calculations as implemented in the OLINDA v1 software with those obtained using ICRP absorbed 11 fractions as implemented in 3D-RD-S, a newly developed software package that implements the MIRD S- 12 value methodology. We also compare ICRP values with a hybrid set of calculations in which alpha-particle 13 energy was assumed completely absorbed in activity containing tissues. the results of Monte Carlo simulations performed using updated and much more realistic voxelized


1
The Medical Internal Radiation Dose Committee published MIRD Pamphlet No. 1 in 1968. This Pamphlet, 1 To study the effects of inhaled radioactivity, the morphological model of respiratory tract is divided into 2 five tissue regions associated with deposited radioactive aerosols along the wall of the respiratory tract.

3
These include ET1 (Anterior Nasal Cavity), ET2 (Posterior Nasal Cavity), BB (Bronchial region), bb 4 (Bronchiolar region), and AI (Alveolar Interstitial) which collectively define the respiratory airways. The updated respiratory tract phantom geometry divides the respiratory tract into several individual source 9 tissue regions into which inhaled radionuclides may deposit. Figure 1 shows the transition from source 10 regions defined in the CE phantom to the morphological phantom developed by the ICRP. The regions of 11 interest for nuclear medicine can be narrowed down to a subset of the regions constituting each section 12 mentioned above. Figure 2 depicts the CE phantom lung regions and the corresponding regions in the ICRP 13 phantom that are relevant to nuclear medicine. As seen by comparing the information on the models in 14 Figure 1 and Figure 2, these are a subset of the comprehensive list of all source regions provided in the There are three regions in the ICRP phantom that are coincident with "lungs" as described in the CE 1 phantom. These ICRP phantom regions are listed as: "lungs," "alveolar-interstitial (AI)," and "lung tissue". The region listed as "lung tissue" corresponds to all soft tissues within the lungs (e.g., lung parenchyma) 6 but exclusive of all blood within the lungs (pulmonary arteries, capillaries, and veins). In contrast, the 7 ICRP region listed as "lungs" includes both the lung tissue and its blood content. This distinction is relevant 8 only for biokinetic modeling, in terms of S values, as these two source regions are identical. Studies cited 9 in ICRP Publication 66 suggest that blood represents about 58% of the mass of the lungs. The "AI" region is defined as the tissue region supporting the terminal bronchioles. It corresponds to the sub-region of the 11 lungs where oxygen and carbon dioxide exchange between air in the lungs and the blood gases take place.

12
This region is particularly relevant to dosimetry associated with inhaled radioactive particles or potentially 13 to radioactive gas released from the circulation and thus a potential critical target region for emissions 14 originating in adjacent source regions. Figure 3 represents the most appropriate target tissue mapping from 15 the CE Phantom to the ICRP phantom. The secretory and basal cells are the relevant targets for radiogenic 16 lung cancer, and thus they are used by the ICRP in their computation of "lung dose" for radioprotection.

17
For RPT, however, a lung target most representative of the risk of lung fibrosis (or other deterministic 18 effect) is more appropriate. The best ICRP target here would be AI. The ICRP's alimentary tract model of Publication 100 [12], and which is implemented in the Publication 5 110 adult phantoms, has additional source regions including the oesophagus, teeth, oral cavity, and salivary 6 glands. A portion of the activity in the oral cavity may also be apportioned to oral mucosa. The "salivary 7 glands" source region includes six regions -the left and right parotid, submandibular, and sublingual 8 glands.

9
The oesophageal wall, oral cavity mucosal lining, teeth surface, and salivary glands are relevant source or 10 target tissues for absorbed dose calculations in nuclear medicine therapy and diagnosis.

11
The colon (or large intestines in the CE nomenclature) regions that are defined in the ICRP phantom differ 12 from those in the CE phantom. To transition from the CE to the ICRP phantom, the total activity (ULI +  Table 1 lists the mass and mapping factor required for an equivalent mapping. Assuming uniform 1 distribution of activity throughout the colon, the mapping factor times the total activity in the large intestines 2 gives the activity in the corresponding ICRP source regions. The ICRP phantom alimentary tract includes target regions deemed important in their impact on overall 5 tissue response to radiation. These include the oral mucosa, and the stem cell layers within the mucosal 6 layers of the stomach, small intestine and large intestine walls. Figure 4 illustrates the corresponding target 7 regions for the CE and ICRP phantoms.

5
The definition of bone endosteum has evolved over time. In the ICRP 30 model, bone endosteum was 6 defined as a single cell layer (10 microns in thickness) along all surfaces of bone trabeculae in trabecular 7 bone, and along the surfaces of the Haversian canals of cortical bone.

8
With the development of the ICRP 110 adult phantoms, the ICRP changed this target tissue (for radiogenic 9 bone cancer risk) to include a 50-micron region of bone marrow along the surfaces of the bone trabeculae, 10 and a 50-micron layer of marrow along the inner bone shafts of the long bones. Cortical bone is no longer 11 considered to house "endosteum" and thus is only a radiation source region and no longer a component of 12 the endosteum target region.

13
To summarize, the endosteum layer in the ICRP phantom and the osteogenic layer in the CE phantom are 14 the corresponding target regions for the skeletal system. The active marrow is another target region common 15 to both the CE and ICRP phantoms. Both phantoms have endosteum as a target -but this target is 1 Figure 5: CE Phantom to ICRP phantom mapping in Skeletal system

2
The spleen, thymus, thyroid, ovaries, testes, uterus, brain, breast and heart are common as source regions 3 in both the ICRP and CE phantoms. In the CE phantom, the heart was divided into its contents and wall.

4
The ICRP phantom considers only the heart wall since total body blood is included as a distinct source 5 region. The lens of the eye, pituitary gland, tonsils, prostate, salivary glands (including the parotid, 6 submandibular and sublingual), teeth, oral cavity, adipose tissue and the ureters are new and additional 7 source regions defined within the ICRP adult phantom.

8
Similarly, there are a few other common target regions included in both models; these are listed in the 9 Appendix. There exist additional target regions corresponding to the additional source tissue in the ICRP 10 adult phantom resulting in a total of 43 target regions.

11
Other source and target regions

12
The CE phantom has a "residual soft tissue (RST)" region corresponding to interior regions of the phantoms 13 that are not already taken by other organs within the body. This RST region is used as a surrogate for 14 "muscle" as a source region and as a target region. In the ICRP phantoms, however explicit geometric 15 models for both "muscle" and "adipose" tissue are defined; there are no undefined regions in the ICRP 2 Target mass differences will impact the absorbed dose comparison.  To facilitate biokinetic modeling wherein a central blood pool is a distinct physiologic compartment, the 9 ICRP phantom considers blood as a distinct source region. Accordingly, source regions correspond to and 10 are assigned the mass of tissue parenchyma. Activity in blood also contributes to the total organ activity,

11
and thus the organ self-doses and cross-doses. Source region activity values estimated from nuclear 12 medicine imaging will include activity in the whole organ (blood + parenchyma). Without special care, 13 using a measured blood value directly would lead to double counting of the blood in such regions since 14 source region activity measured from the image includes both the radiopharmaceutical activity in the organ 15 parenchyma as well as the radiopharmaceutical activity within the blood pool of that same organ. This can 16 be avoided by subtracting the fraction of total body blood activity that is localized within the blood pool of

11
The particle yields and energies used for the absorbed dose calculations were obtained from ICRP

5
The time-integrated activity coefficient (TIAC = residence time), i.e., the integral over time of the activity 6 in the organ divided by the injected activity, expressed as fraction of injected activity [hours per g (FIA-7 h/g)] for each pre-clinical organ are shown in Table 4.
8  The TIAC of the pre-clinical "core bone sample of the distal femur" was assigned to human cortical bone 2 and "L2 vertebral body" was assigned to trabeculae bone using equation 1. This assignment was used to 3 obtain electron and photon dose contributions to the osteogenic cell target in the CE phantom region. The 4 L2 vertebral body measurement was also used in the hybrid calculations to apportion the TIAC to 5 osteogenic cells, which was taken to be equivalent to the trabecular bone endosteum layer as defined in 6 [13]. In the calculations using ICRP data, trabecular bone surface (which has zero mass in the ICRP model) 7 serves as a source region. Thus, we assigned the TIAC to this region using the ICRP trabecular bone marrow 8 mass in the conversion from pre-clinical L2 TIAC to human TIAC.

9
The large intestine TIAC was apportioned to human upper and lower large intestine for the CEP and Hybrid 10 calculation. For the ICRP calculation, the adult reference male phantom was used and the large intestine

11
TIAC was apportioned to the right, left and rectosigmoid colon regions based on the mapping factors 12 described above. OLINDA-based estimates of the alpha-particle dose to each target tissue were replaced by separate 4 calculations that used the summed alpha-particle energy for thorium-227 and radium-223 as listed in Table   5 3. The equations describing this approach are given below.
, and (4) D e,ph (r T ) is the electron (e) and photon (ph) dose contribution to target region, r T , 11 D α (r T ) is the alpha-particle (a) dose contribution to r T , 12 D(r T ) is the total absorbed dose to r T , 13 S e,ph (r T ← r S ) is the S value, i.e., the absorbed dose to r T from e and ph emissions per nuclear 14 transition in source region, r S , is the mean energy emitted per nuclear transition of the i th alpha particle (α) 17 emission, and 18 ϕ(r T ← r S ; E i α ) is the fraction of energy emitted per nuclear transition in the source region, r S , that is Since the calculations are performed at the level of whole organ dimensions, the following holds: 2 Accordingly, equation 4 reduces to: 4 where, 5 Δ α is the total α-particle energy emitted per nuclear transition of the radionuclide, 6 (Gy-kg)/(Bq-s).

7
Equation 7 was derived assuming the emitted α-particle energy is completely absorbed in r T .

8
There is ample evidence that the first assumption (Eq. 6) fails for some tissues. In particular, the distribution 9 of the mAb through the kidneys is known to be non-uniform. The second assumption (Eq. 7) is met if the target region is substantially larger than the 60 to 100 µm range of the α-particles. Since radium-223 has a 11 biodistribution different from thorium-227 labeled anti-HER2-antibody, Equation 7 was implemented 12 separately for these two α-emitters.

16
[13], were used to calculate S values for the trabecular active marrow (TAM) and the trabecular bone endosteum (TBE) from alpha-particle emissions in these two volumes as follows: 19 , and (11) 1 where, 2 M(TAM) and M(TBE) are the mases of TAM and TBE, respectively, and 3 E i α and Y i α are the energy and yield, respectively, of alpha-particle i emitted by thorium-227 or radium-4 223 and its daughters.

5
The TAM and TBE masses (Table 5) were obtained from data in ICRP 89, ICRP 70 and Watchman, et al.
6 [10,13,14]. The mass of the TBE was obtained as the product of the trabecular bone surface area, the 10 7 µm-thickness of the endosteal layer, and the density of the endosteal layer. The mass used for the TAM 8 was consistent with the cellularity of the reference 40-year-old male. 9
3 CEP and hybrid do not distinguish between wall and contents for small intestine region.

Absorbed Doses
1 Table 8 through 11 list, respectively, the photon, beta particle, α-particle and total specific absorbed dose 2 calculated using the three different methods. The photon, beta particle, α-particle and total specific absorbed 3 doses are plotted in Figure 7-11, respectively. 4    Figure 8: Comparison between the sum of beta-particle and electron absorbed doses calculated using the CEP,

2
Hybrid and ICRP methods.
3  Phantoms, and modifications thereof, as implemented in OLINDA v1). The absorbed fractions calculated 5 using these phantoms focused primarily on photon transport; the dose contribution from electrons was set 6 either to 1 (organ self-dose) or zero (cross-organ dose). The absorbed dose contribution to wall from 7 contents was set to 0.5 of the content self-irradiation absorbed dose. To address the increasing use of alpha-8 particle emitting radiopharmaceuticals for therapy, absorbed fractions for alpha-particles were calculated 9 and incorporated into S-values for alpha-particle emitters that are used in OLINDA. In 2017, the ICRP released new and much more realistic phantoms [18]. These were generated by segmenting patient CT 11 scans. The specific absorbed fractions calculated using these new phantom geometries include both electron 12 and alpha-particle transport (using stylized models for selected tissue). As Figure 11 shows, there are 13 substantial differences in anatomical realism between the two phantom types. Despite the better anatomical representation and improved accuracy, calculations using these new phantoms 1 have not been widely adopted. The comparisons provided in this work are intended to help transition the 2 field to these new models, which are anatomically and computationally more accurate. As a first step, we 3 established a correspondence between CEP and ICRP phantom source/target tissue nomenclature ( Figure   4 1-6). We have also endeavored to identify tissue mass differences and use these to provide guidance on 5 how to apportion TIAC originally derived for CEP anatomic geometry (Table 1, Appendix).

6
A B C 1 differ in how alpha particle absorbed dose is calculated. This was done because the provenance of alpha-2 particle S values used in OLINDA has not been established. Accordingly, the second methodology (hybrid) 3 explicitly describes how alpha particle energy was apportioned in the calculations, making it possible to 4 understand and interpret differences in the alpha-particle contributions to the absorbed dose. In this 5 approach, we assumed complete absorption of alpha-particle energies for all tissues except those associated 6 with the skeleton. To calculate alpha-particle absorbed dose for skeletal regions (i.e., the red marrow and 7 trabecular bone surface), we used alpha-particle absorbed fractions published by Watchman,et al. [13]).
by thorium-227 and by radium-223 and its daughters. Furthermore, in the latest ICRP skeletal bone 11 geometry, the osteogenic bone surface has been replaced by a 50-µm thick layer of endosteal cells, referred 12 to as the trabecular bone endosteum. energy of alpha particles (with a yield greater than 20%) and corresponding alpha emitters is shown by 1 arrows. TAM <-Trabecular Active Marrow, TBE <-Trabecular Bone Endosteum.

2
The energies of the alpha particles emitted by thorium-227 and its daughters are between 5.5 and 7.5 MeV.

3
As demonstrated by the arrows in Figure 12, the corresponding skeletal average absorbed fraction for 4 decays originating in the trabecular bone surface (previously referred to as the trabecular bone endosteum 5 (TBE)) irradiating the trabecular active marrow (TAM) range from 0.20 to 0.22.

6
The data set used for the absorbed dose comparisons was for a therapeutic alpha-particle emitting 7 radionuclide ( Figure 6 and Table 3) with measurements obtained from counting of extracted pre-clinical 8 tissues. Pre-clinical tissue TIAC values were provided as input for the calculations (Table 4). This required 9 conversion of TIAC from pre-clinical to human (Eq. 1).

11
The values in the table reflect a number of decisions made across the different methods. The first 12 observation is that the TIAC assigned to the same tissue differs across the different methods. For example,

13
the TIAC for thorium-227 in the adrenals was 0.21 h for the CEP/Hybrid method but was 0.226 h for the 14 ICRP method. This is because the translation of pre-clinical TIAC to human TIAC used ICRP phantom 15 organ and whole-body masses (Eqn. 1) instead of the CE/OLINDA phantom masses.

18
TIACs for these tissues were treated as sources in the respective contents of each organ. For the large 19 intestine, the TIAC was apportioned to human equivalent upper and lower large intestine contents based on 20 the fractional mass of the walls. In the hybrid calculations, we placed the TIAC into the wall assuming 21 complete absorption of alpha-particle energy in the wall using the mass of the wall listed for the CE phantom (Table 1 of the Appendix). The ICRP phantom provides both the wall and contents as source regions; we 23 placed the TIAC in the wall: since the measurements came from direct sampling, we assumed that the wall 24 was counted and not the contents. As indicated in Error! Reference source not found., the digestive tract anatomy is redefined in the ICRP phantoms: the upper and lower large intestines have been replaced by the 1 right and left colon and the rectosigmoid colon. Accordingly, the TIACs used for these tissues in the ICRP 2 phantom were apportioned as shown in Table 2.
3 Table 7 shows that the ICRP phantom includes tissues not available in the CEP. The data set used in this 4 work included TIAC values for the mesenteric lymph nodes and prostate. These were assigned to the extra 5 thoracic lymph nodes and prostate, respectively. The source tissues in the ICRP phantom are indicated in 6

7
To understand the impact of adopting ICRP-phantom-derived S-values, the individual absorbed dose 8 contribution for each emission type to the total absorbed dose is listed for each of the three methods (Table   9 8-10 and Figure 7-10). Since both thorium-227 and radium-223 emit alpha-particles, the total absorbed dose 10 across all daughter emissions (Table 11 and Figure 10) is dominated by the dose contribution arising from 11 alpha-particles.

12
In Table 8-10, differences in target nomenclature are indicated. For example, "Lungs" in the CEP phantom 13 corresponds to "AI" in the ICRP phantom notation (Figure 3). Likewise, "osteogenic cells" correspond to

20
The percent difference in absorbed dose for photon, electron and alpha-particle emissions arising from 21 thorium-227 are listed on  prostate n/a n/a n/a n/a r-marrow 58.6 50.9 -212.8 -37.9 LLI-wall (CEP, hybrid)/rc-stem (ICRP) 92.4 n/a n/a n/a rs-stem n/a n/a n/a n/a si-wall (hybrid)/si-stem (ICRP) 50.8 n/a n/a n/a In the ICRP method, photon absorbed dose for every tissue listed is greater than that calculated using the 4 CEP method. This is most likely a combination of greater accuracy in the absorbed fraction calculations 5 due to substantially more powerful computing capabilities and also anatomical differences in the phantoms 6 ( Figure 11). The former probably better accounts for absorption of low energy photons, while the latter 7 provides a contiguous anatomy with no gaps between organs. Another possible source of differences could 8 be due to differences in the decay spectrum used for thorium-227. for self-dose and zero for cross-organ doses, the substantial differences are likely due to differences in the 7 electron spectrum used for the electron absorbed dose calculations. Interestingly, there is much better 8 agreement in the electron dose between the OLINDA implementation of CEP and that obtained using ICRP 9 for 177 Lu (data not shown). Lutetium-177 is primarily a beta-particle emitter; the beta-particle energy emitted per disintegration of 177 Lu is almost 3 orders of magnitude greater than that for monoenergetic 11 (Auger, conversion) electrons. Thorium-227 does not emit beta-particles, and the monoenergetic electron 12 energy emitted is predominantly due to emissions below 1 MeV. Accordingly, the difference in electron 13 absorbed doses may be due to differential handling of low energy emissions in the Monte Carlo simulations 14 or possibly a decay scheme that abridges the monoenergetic electrons for thorium-227 used in the absorbed 15 fraction calculations in the OLINDA implementation of the CEP phantoms.

16
In evaluating the differences in alpha-particle absorbed dose, we confined the discussion to the Hybrid 17 calculation where the methodology and input data used to arrive at the dose estimates are well described.

18
The percent differences between the hybrid and ICRP results for alpha particle absorbed dose are far less 19 for most organs than that observed for photon and electron absorbed doses. Since the absorbed dose from 20 thorium-227 is dominated by its alpha-particle emissions, the discrepancy in photon and electron absorbed 21 doses is practically inconsequential. The brain, osteogenic cells, gallbladder wall and red marrow show differences greater than 5%. These are due to a combination of differences in tissue mass/tissue definition 23 and the TIAC assigned for the CEP/hybrid combination compared to the ICRP values. The very small percent differences in other tissues are likely due to rounding errors associated with mass and total alpha-1 particle energy used in the calculations 2 The discrepancy between alpha absorbed dose for the gall bladder wall arises because complete absorption 3 of alpha-particle energy was assumed in the hybrid calculations but not the ICRP calculations. For example, 4 the results suggest that approximately 40% of the alpha-particle emissions arising from decays in the gall 5 bladder wall deposit energy outside the wall. The large percent differences in absorbed dose for osteogenic 6 cells and for the red marrow are examined in Table 13. As indicated in the methods section, the RM and 7 osteogenic cell absorbed doses using the hybrid method were calculated using cross-tissue specific absorbed 8 fractions to calculate the S-values shown in Table 6. the RM-to-RM absorbed fraction obtained using the ICRP vs hybrid methods. Applying this same 14 calculation for the endosteal (or osteogenic cell) target, the ICRP-equivalent absorbed dose to bone calculations is the same, these differences are probably due to a combination of differences due to tissue geometry and Monte Carlo techniques. Comparing these self-dose % differences with those listed on Table   18 12 suggests that the values in Table 12 for these tissues are dominated by differences in the self-absorbed

11
HH is an employee and shareholder of Bayer AG.

12
SJ is an employee of Bayer AG.

13
TP is an employee and shareholder of Bayer AG.
14 AC is an employee and shareholder of Bayer AG.

15
All other authors declare that they have no competing interests.

16
Funding and Acknowledgements: The work described in this paper is funded by SBIR grant 1 R43 participated in the analysis and discussion. TP, HH, AP and SJ were responsible for data collection. WB