Phoswich scintillation counters are used to detect actinides deposited in the lungs. The resulting spectra, however, contain Compton background from the decay of 40K, which occurs naturally in the striated muscle tissue of the body. To determine the counts due to actinides in a lung count spectrum, the counts due to 40K scatter must first be subtracted out. The 40K background in the phoswich NaI(TI) spectrum was predicted from an energy region of interest called the monitor region (95–129 keV), which is above the 238Pu region (11–32 keV) and the 241Am region (40–82 keV), where photopeaks from 238Pu and 241Am occur. Empirical models were developed to predict the backgrounds in the 238 Pu and 241Am regions by testing multiple linear and nonlinear regression models. The initial multiple regression models contain a monitor region variable as well as the variables gender, (weight/height)α, and interaction terms. Data were collected from 64 male and 63 female subjects with no internal exposure. For the 238Pu region, the only significant predictor was found to be the monitor region. For the 241Am region, the monitor region was found to have the greatest effect on prediction, while gender was significant only when weight/height was included in a model. Gender-specific models were thus developed. The empirical models for the 241Am region that contain weight/height were shown to have the best coefficients of determination (R2) and the lowest mean squares for error (MSE).