A Study of Total Measurement Error in Tomographic Gamma Scanning to Assay Nuclear Material with Emphasis on a Bias Issue for Low-Activity Samples

Field experience with the tomographic gamma scanner to assay nuclear material suggests that the analysis techniques can significantly impact the assay uncertainty. For example, currently implemented image reconstruction methods exhibit a positive bias for low-activity samples. Preliminary studies indicate that bias reduction could be achieved at the expense of increased random error variance. In this paper, we examine three possible bias sources: (1) measurement error in the estimated transmission matrix, (2) the positivity constraint on the estimated mass of nuclear material, and (3) “improper” treatment of the measurement error structure. We present results from many small-scale simulation studies to examine this bias/variance tradeoff for a few image reconstruction methods in the presence of the three possible bias sources.