^{U.S. Transuranium and Uranium Registries}

_{Conference Contributions}

## 67^{th} Annual Meeting of the Radiation Research Society, San Juan, Puerto Rico, October 3-6, 2021

*Due to the COVID-19 pandemic, the Radiation Research Society meeting was moved to a virtual platform, and the USTUR’s posters were withdrawn.*

USTUR researchers look forward to participating in the 67^{th} Annual Meeting of the Radiation Research Society, which will be held October 3-6, 2021. Two posters will focus on the use of latent variables to estimate the concentration of plutonium in the human skeleton from the concentrations in a few bones. A third poster will discuss the challenging task of associating compartments in biokinetic models with the source regions that are used to calculate specific absorbed fractions (SAFs), and therefore dose.

Perils of Internal Dosimetry: Linking Biokinetic and Dosimetric Models

Martin Šefl (USTUR), Maia Avtandilashvili (USTUR), Stacey L. McComish (USTUR), Joey Y. Zhou (DOE), Sergei Y. Tolmachev (USTUR)

Radiation dose assessment for internally deposited radionuclides requires a combination of biokinetic and dosimetric models. For example, a biokinetic model of inhaled plutonium is described by the International Commission on Radiological Protection human respiratory tract, human alimentary tract, and plutonium systemic models. Calculations of absorbed doses to various organs are simplified by using specific absorbed fractions (SAF). A SAF is the fraction of the particle energy emitted in a source region deposited in a target region per mass of the target tissue. The calculation of the dose then appears straightforward. For plutonium, the absorbed dose in the target region equals *D*_{t} = Σ_{s} *N*_{s}×*E*_{s}×*SAF*(t←s) where *N*_{s} is the number of disintegrations in a source region, *E*_{s} is the total alpha energy emitted per one disintegration in the source region, and *SAF*(t←s) is the specific absorbed fraction for a combination of source-target regions. Conversely, in a biokinetic model, compartments are not always the same as the source regions defined for the *SAF*. To calculate the number of disintegrations in a source region, the biokinetic compartments must be associated with the source regions. For some organs, the process is straightforward; for others, it is not. For example, 59 compartments are needed to model plutonium inhalation, and there are 79 source regions (tissues) for which the *SAF*s are defined. The associations depend on a specific biokinetic model, and, unfortunately, biokinetic models are not always accompanied by a guide on how to link the biokinetic model compartments with the corresponding dosimetric source regions. This presentation describes a step-by-step association guide for plutonium inhalation. We suggest that an explicit association guide should be provided for all radionuclides to avoid ambiguity and confusion. [USTUR-0589-21A]

Plutonium in Human Skeleton: Latent Bone Modeling

Sergei Y. Tolmachev (USTUR), Maia Avtandilashvili (USTUR), George Tabatadze (USTUR), Joey Y. Zhou (DOE)

The skeleton is a major depository site for plutonium in the human body. In radiation protection, a long-term standing question is: What is the most accurate and precise way to estimate the skeleton plutonium concentration and activity from the analysis of a limited set of bones? To answer this question, several approaches were used including a multiple linear regression (MLR) analysis. The key limitation of the MRA approach is multicollinearity among independent variables, since the activity concentrations from individual bones are highly correlated, resulting in unstable and imprecise estimates of model coefficients. This study applied principal components regression (PCR) to the data from 19 United States Transuranium and Uranium Registries whole-body tissue donors by performing principal components analysis (PCA). This was followed by the selection of a set of 1 to 3 principal components as latent bones (independent variables) for a subsequent multiple linear regression modeling. Total skeleton plutonium activity concentrations (*C*_{skel}) for these individuals were estimated based on post-mortem radiochemical analyses of bone samples (*C*_{bone}) from the right side of the skeleton. The PCAs show that a single latent bone alone (the first principal component) explains over 95% of the total variance due to a high correlation of activity concentrations among individual bones. A Monte Carlo method was developed to evaluate uncertainty in latent bone model (LBM) estimates of *C*_{skel} and the effect of osteoporosis on LBM was evaluated. The LBM approach was also applied to select the best combination of sample bones for estimating the *C*_{skel}. [USTUR-0588-21A]

Comparison of Latent Bone Modeling and Simple Average Method for Estimating Plutonium Activity Concentration in Human Skeleton

Joey Y. Zhou (DOE), George Tabatadze (USTUR), Maia Avtandilashvili (USTUR), Sergei Y. Tolmachev (USTUR)

The United States Transuranium and Uranium Registries (USTUR) holds data from 290 partial-body tissue donors for whom only 2 to 8 bone samples were collected at autopsy and radiochemically analyzed for plutonium. The method currently used at the USTUR to estimate skeleton plutonium activity concentrations, simple average method (SAM), implies that the arithmetic average of collected bone sample concentrations represents the total skeleton concentration. The USTUR also holds data from 14 non-osteoporotic whole-body donors for whom ‘true’ values of total skeleton plutonium activity concentrations were estimated based on radiochemical analyses of all measured bones from the right side of the skeleton. A recently developed latent bone model (LBM) applies principal components regression (PCR) to reduce uncertainties in plutonium activity concentration estimates from measurements of a limited set of bone samples. This study used a simulation approach to compare accuracy and precision of the LMB and SAM. The analytical bone dataset consisted of measured plutonium concentrations in up to 90 individual bones from 14 non-osteoporotic whole-body tissue donors. For each simulation run, Root Mean Square Errors (RMSEs) were determined for the LBM and SAM, and 10,000 simulations were run for a given number of individual bones (2 to 5). The distributions (mean and standard deviation) of RMSEs obtained from the 10,000 simulations were used to compare accuracy and precision of the LBM and SAM. The results showed that the LBM approach significantly improved the total skeleton concentration estimates. The relative mean (accuracy) reductions of the LBM vs. SAM were 55.5%, 57.4%, 59.1%, and 60.4%; relative standard deviation (precision) reductions were 65.2%, 66.6%, 67.9%, and 68.6% for 2, 3, 4, and 5 bone samples, respectively. [USTUR-0590-21A]