Purpose: In x-ray scatter projection imaging, an innovative technique that yields higher tissue contrast, a pencil beam is raster scanned over the object to generate scatter data. Using multiple beams reduces acquisition time but the individual scatter signals overlap at the detector. We have used a maximum likelihood expectation maximization (MLEM) algorithm to disentangle the scatter signals. For objects with highly-attenuating (e.g. bone) or minimally-scattering (e.g. air) materials, the MLEM algorithm performs poorly and the scatter image has ghosting artifacts. We propose the use of thin vanes between the object and detector to eliminate scatter pattern overlap.
Methods: In a simulation, five parallel coplanar 33.2 keV pencil beams were used to image a 2 cm thick phantom comprising fat, muscle, bone, and air. At this energy, 3.50Â° â‰¤ Î¸ â‰¤ 5.57Â° scatter is used to maximize fat-muscle contrast. A flat-panel detector 20 cm behind the phantom recorded the Laue ring scatter pattern. A planar, 0.2 mm thick, W/Cu alloy vane was placed on edge between each adjacent beam pair. Scatter pattern acquisition with/without vanes was simulated for phantom incident air kerma over the range 0.03â€“10 mGy. Using a single beam, perfectly-disentangled non-vane scatter patterns were also simulated.
Results: There was noticeable ghosting in MLEM-reconstructed images but none in images acquired with vanes. For the planar vane shape, 42.0% of the 3.50Â° â‰¤ Î¸ â‰¤ 5.57Â° scatter from the three middle beams was admitted. For all air kerma, the fat-muscle contrast-to-noise ratio (CNR) in the vane images was confirmed to be approximately equal to the square root of the fraction of scatter admitted by the vanes times the CNR in the perfectly-disentangled non-vane images.
Conclusion: The pre-detector vanes eliminate ghosting artifacts but reduce CNR by blocking scattered photons. Curved vanes would admit more scatter from the middle beams, increasing CNR.
Funding Support, Disclosures, and Conflict of Interest: This work was supported by the Natural Sciences and Engineering Research Council of Canada and by an Ontario Graduate Scholarship.