X-ray tomography of extended objects: a comparison of data acquisition approaches

Abstract

The penetration power of x rays allows one to image large objects, while their short wavelength allows for high spatial resolution. As a result, with synchrotron sources, one has the potential to obtain tomographic images of centimeter-sized specimens with sub-micrometer pixel sizes. However, limitations on beam and detector size make it difficult to acquire such data of this sort in a single take, necessitating strategies for combining data from multiple regions. One strategy is to acquire a tiled set of local tomograms by rotating the specimen around each of the local tomogram center positions. Another strategy, sinogram oriented acquisition, involves the collection of projections at multiple offset positions from the rotation axis followed by data merging and reconstruction. We have carried out a simulation study to compare these two approaches in terms of radiation dose applied to the specimen, and reconstructed image quality. Local tomography acquisition involves an easier data alignment problem, and immediate viewing of subregions before the entire dataset has been acquired. Sinogram oriented acquisition involves a more difficult data assembly and alignment procedure, and it is more sensitive to accumulative registration error. However, sinogram oriented acquisition is more dose efficient, involves fewer translation motions of the object, and avoids certain artifacts of local tomography.