Abstract
Iterative tomographic reconstruction has been established as a viable alternative for data analysis in phase-sensitive x-ray imaging based on the edge-illumination principle. However, previously published approaches did not account for drifts of optical elements during a scan, which can lead to artefacts. Up to now, the strategy to reduce these artefacts was to acquire additional intermediate flat field images, which were used to correct the sinograms. Here, we expand the theoretical model to take these effects into account and demonstrate a significant reduction of (ring)-artefacts in the final reconstructions, while allowing for a significant reduction of scan time and dose. We further improve the model by including the capability to reconstruct combined absorption and phase contrast slices, which we experimentally demonstrate to deliver improved contrast to noise ratios compared to previously employed single shot approaches.
Highlights
INTRODUCTIONCompared to conventional x-ray radiography, which provides absorption contrast only, refraction sensitive x-ray imaging methods deliver phase and absorption contrasts simultaneously[1 ]
Iterative tomographic reconstruction has been established as a viable alternative for data analysis in phasesensitive x-ray imaging based on the edge-illumination principle
This method, called computed tomography (CT), is a staple in current diagnostic imaging and the refraction sensitive techniques mentioned above have been successfully combined with tomography[15,17,18]
Summary
Compared to conventional x-ray radiography, which provides absorption contrast only, refraction sensitive x-ray imaging methods deliver phase and absorption contrasts simultaneously[1 ]. While the utilised assumption of proportionality between absorption and phase contrast is strictly true only in the case of a single homogeneous material present in the beam and precludes a quantitative interpretation for more complex material systems, in practice qualitative interpretation of the images such as morphological information (e.g. pore size distribution) is frequently of interest This assumption has been demonstrated to improve image quality for EI in some cases[33] and simplifies the iterative reconstruction as the number of retrieved values are only half compared to the previously published method[25 ]. The iteration was carried out until satisfying convergence was reached, which took 58 iteration steps and 93 s on a single core of a standard modern desktop PC
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have