Characterizing a Tomographic Gamma Scanner

The Tomographic Gamma Scanner (TGS) uses high-resolution gamma spectrometry to nondestructively assay radioactive items. Attenuation corrections are based on the joint application of linear attenuation coefficient and emission images, derived from the scan data, using ray-tracing methods on a stack of volume elements. In principle the technique is a very powerful one although care must be taken during design and implementation to ensure accuracy. The image reconstruction algorithms are also quite sensitive to the statistical quality of the input data. The TGS places greater demands on hardware than traditional non-imaging gamma-scanning instruments. Mechanical motions and alignments are more critical and must be controlled more precisely and accurately. In addition data taking rates (of data grabs or views) and dead time corrections are more challenging. We have recently developed a new TGS instrument to best industrial practices. In this paper we describe some of the exacting performance tests undertaken to confirm that the operation was in accord with design objectives and physics limitations. These tests included measuring the fundamental characteristics of the scanner such as the point spread spatial response function. Comparisons were made against the TGS analysis model. Careful measurements were performed to demonstrate that the movements of the scanning components were properly aligned and that the range and pace of motion could be controlled accurately. By deliberately misaligning the system, the ability to predict the magnitude of the effect and to compensate for the misalignment was examined using TGS simulations. We confirmed that complete two pass assay times could be accomplished reliably using assay times as short as 20min or as long as many hours. Sinogram plots using point and rod sources were taken to show directly that reproducibility and timing uncertainty in the data synchronization over the full operating range. Difficult to measure items, for instance dense heterogeneous drums, tend to amplify image reconstruction bias. We investigated this effect by acquiring transmission and emission data over different durations (0.5 - 20 hours) for several test drums and analyzing the data using all time pairings. This provides the context about which to give the characteristic response data practical meaning. In other words, only when the magnitude of the potential alignment, motion and rate dependent errors are known in relation to the overall uncertainty, including those contributions that are inherent to the TGS method, can one make well informed decisions about how well they need to be controlled in the field. We discuss general ways of approaching difficult to assay cases so that the results remain robust and bounded by the total measurement uncertainty estimate.