Results and error analysis of a reference voxel normalization method for neutron computed tomography partial volume voxel water quantification

Although water quantification using neutron imaging has been successfully applied to water distribution studies of hydrogen-powered fuel cells, the technique produces two-dimensional images, which makes it difficult to determine the exact "depth" inside of an object where an imaged mass of water is located. In contrast, neutron computed tomography (NCT) is an imaging technique that generates a three-dimensional volumetric reconstruction of an object's interior geometry, allowing determination of a structure's exact spatial location. In a reconstruction, determination of the fractional water masses present in partial volume voxels is required for an accurate quantification. Such a NCT partial volume voxel water quantification technique has been developed at The Pennsylvania State University Radiation Science and Engineering Center (RSEC). Several approaches with water-filled cylindrical aluminum samples were used to test the technique: MCNP simulations, samples 30 mm from the detector, and samples 140 mm from the detector. Quantification results were within 0.2% of the theoretical for the MCNP simulations and within 2% for the samples 30 mm from the detector. Significant error, as large as 47%, was seen with the samples 140 mm from the detector. Geometric un-sharpness was determined to be the dominant source of this error. Comparisons between capturing a full set of projections, called true-CT, and a single projection that is duplicated to approximate a full set, called pseudo-CT, were also performed. Results showed pseudo-CT well approximated true-CT, within 1%, for a sample with geometry simple enough that its attenuation map was the same for every viewing angle.