The development of models to predict the in vivo background in phoswich lung count spectra

Author

Nancy W. Richards

Date of Award

8-1995

Degree Type

Thesis

Degree Name

Master of Science

Major

Nuclear Engineering

Major Professor

Laurence F. Miller

Committee Members

Darryl J. Downing, James E. Turner, Gloria T. Mei, James R. Watts, Govind R. Rao

Abstract

Lung count spectra acquired with phoswich detectors are dominated by Compton-scattered photons from the decay of natural potassium-40 (K-40) that is distributed in the body. To detect the presence of actinides in the lungs, the K-40 background contribution must be estimated and subtracted out. Data analysis at the Oak Ridge National Laboratory currently employs a spectral region called the monitor region (129 - 166 keV) to predict the backgrounds in the ²⁴¹Am (40 - 82 keV) and ²³⁸Pu (11 - 32 keV) regions. This study refines the current methodology. Prediction models are developed from multiple linear and nonlinear regression models that contain the biometric variables, gender and (weight/height)”, and a second spectral region called the monitor! region at 95 - 129 keV. Data are collected from 64 male and 63 female subjects free of internally deposited activity. Interaction terms like “(weight/height)” x monitor2 region x gender" are also included in the analysis. For the ²⁴¹Am-region models, all variables are statistically significant, but the monitor2 region has the greatest effect. Gender is significant only when weight/height is included, as higher weight/height ratios are associated with males, and lower weight/height ratios, with females. For the ²⁴¹Am region, gender-specific models that include the monitor2-region and weight/height variables have the highest coefficients of determination (males: R² = 0.89; females: R² = 0.93) and the lowest mean squares for error. For the ²³⁸Pu-region model, the R² is 0.78, and the only significant predictor is the monitor2 region (p < 0.0001), as the biometric variables have little influence. The monitor2 and the monitor regions do not differ significantly in their ability to predict the background in the ²³⁸Pu region. Recommendations are as follow: 1) Use the monitor2-region and weight/height variables in the ²⁴¹Am-region models; 2) and calculate the error of the predicted value and studentized residuals for all models.

Recommended Citation

Richards, Nancy W., "The development of models to predict the in vivo background in phoswich lung count spectra. " Master's Thesis, University of Tennessee, 1995.
https://trace.tennessee.edu/utk_gradthes/11248