Emerging from the field of quantum optics and initially believed to be underpinned by quantum-mechanical "spooky action at a distance", the field of ghost imaging has rapidly achieved prominence in studies using classical visible light [1].
This form of imaging is counter-intuitive. In ghost imaging, photons from a source pass through a speckle-making mask, leading to a spatially random pattern A being measured over the surface of a position-sensitive detector. A beam-splitter then removes a very small fraction of the photons, which pass through an object and are then recorded by a single-pixel "bucket" detector that merely records the total number B of photons falling upon it. This process is repeated for a number of different mask positions. While no photon that ever passes through the object is ever registered by a position-sensitive detector, and no photons measured by the position sensitive detector ever pass through the object, the correlation between A and B can be used to reconstruct the object [1].
Ghost imaging using x-rays was only very recently achieved, with the first proofs of concept for one-dimensional x-ray ghost imaging being published by Yu et al. [2] and Pelliccia et al. [3]. This was soon extended to x-ray ghost imaging of two-dimensional objects [4,5]. Finally, based on the theory and computer modelling of Kingston et al. [6], the first experimental realisation of ghost tomography (using potentially any form of radiation, not just x-rays) was reported by Kingston et al. [7]. The experimental setup was as described above, but with the additional feature that the sample was rotated to a number of different angular orientations.
We discuss the origins of ghost imaging, explain the key principles underpinning the method, review the current state of art in x-ray ghost imaging in 1D (line scans), 2D (radiography) and 3D (tomography), consider some key drivers such as the quest for ever-reduced dose, and speculate regarding future developments. We attempt to reduce the counter-intuitive nature of the method to a retrospectively obvious simplicity, and address the obvious question of: "Why would one want to perform tomographic imaging in this peculiar manner?"
[1] Katz et al., APL 95, 131110, 2009.
[2] Yu et al., PRL 117, 113901, 2016.
[3] Pelliccia et al., PRL 117, 113902, 2016.
[4] Zhang et al., Optica 5, 374-377, 2018.
[5] Pelliccia et al., IUCrJ 5, 428-438, 2018.
[6] Kingston et al., IEEE Trans. Comput. Imag. 5, 136-149, 2019.
[7] Kingston et al., Optica 5, 1516-1520, 2018.