Multiview Positron Attenuation Tomography

Positron attenuation tomography (PAT) is a new modality for imaging the linear attenuation coefficients (LACs) of magnetically constrained β⁺-ray beams in matter. A complete 3-D LAC image of an object can be made by measuring the positron annihilation rate density within it at a single orientation relative to the beam, and then applying the PAT transformation. The spatial resolution of this image is limited by the resolution of the positron emission tomography (PET), or other, system used to acquire the data as well as the scattering and gyration of the positrons around the magnetic field lines. The finite resolution of the PET system also leads to nonlinear artifacts associated with extended LAC discontinuities parallel to the beam. The mass thickness of an object that can be imaged is limited by the positron beam's range in it. However, due to the directional nature of the imaging process as well as its discrete sampling, PAT images acquired at different view angles may each carry unique information on the object's structure. This paper describes a forward model of PAT image formation and a backpropagation algorithm that, when used together to iteratively combine the data from multiple views, can improve the spatial resolution of PAT LAC images, extend their effective field of view, and reduce artifacts. The technique is demonstrated on measured and simulated data.