Measurement of the physical PSF for an integrated PET/MR using targeted positron beams

In a uniform, static, 3T magnetic field, positrons emitted by a radioactive source such as 18F or 68Ga are constrained by the Lorentz force to follow a helical path of about a millimeter in radius along the B field lines. In air, these positrons may have an axial range of up to a few meters however, and thus form positron "beams". Intense annihilation sources a few tenths of a millimeter thick can be produced by cutting such beams with thin targets. Due to their high spatial resolution and minimal positron range effects, such sources can be used to make precise measurements of the physical point-spread function (pPSF) of the PET detectors in an integrated PETIMR system, such as the 3T Biograph mMR (Siemens Healthcare), that has its PET component interior to the MR magnet. In this paper we describe measurements of the pPSF of the mMR using this targeted positron beam technique. We find the shape of the kernels is quasi-Gaussian, but with longer tails. The FWHM of the axial component of the kernel is 3.35 mm, with no dependence on axial or radial position in the gantry. The width of the radial component varies from 3.5 mm at the center of the field of view (FOV) to 7.5 mm at 25 cm radial offset, with no dependence on axial position. As we have have previously reported, the pPSF needs to be modified to account for numerical broadening effects in order to produce the most accurate iterative PSF reconstructions. We describe a new methodology for determining this effective PSF (ePSF), based on imaging syringe sources of known diameters. We find the ePSF kernel has a FWHM of 2.2 mm at the center of the FOV. We compare no-PSF, ePSF and pPSF-based reconstructions on both phantom and human studies. We find that the ePSF kernel reduces hyper-resolution and edgeovershoot artifacts relative to the pPSF kernel, although with less s