Positron Range Modeling for Low-Dose Rb-82 Cardiac PET

The resolution degradation in PET is usually compensated by incorporating point spread function (PSF) kernels in the reconstruction system matrix. These PSF kernels can be either simulated using Monte Carlo simulations or measured using F-18 point sources in a real scanner. High energy emitters including Rb-82, however, have long positron ranges which can easily exceed ~5mm FWHM depending on the medium, and therefore causing relatively poor image resolution when using kernels measured by F-18 point sources. The positron range of Rb-82 is conventionally modeled as a local, spatially invariant kernel generated by analytical models or Monte Carlo simulations. However, these approaches do not include spatially varying nature of resolution degradation in a real PET system. In this study, we propose to decompose the PSF kernel into two components, one based on the spatially variant F-18 point source measurements taken at multiple locations in a real scanner and a second component that is a spatially invariant differential kernel that captures the differences between the analytical F-18 and Rb-82 kernels. The resulting kernel is approximated by a parameterized 3D asymmetric Gaussian function. We compared the proposed kernels with the measured F-18 point source kernels on five rest/stress myocardial perfusion studies acquired with either 25mCi or 30mCi Rb-82 infusion on a SiPM-based PET/CT scanner. The results show that incorporating the positron range in the PSF kernels improved the image resolution and myocardium-to-blood pool contrast with lower background noise.