Stepwise linear regression modeling of the point spread functions of a multi-pinhole SPECT camera for 1-123 datscan imaging

We investigate PSF modeling strategies for clinical 1-123 DaTscan imaging on a dual-headed mixed collimator SPECT system consisting of fan and multi-pinhole (MPH) collimators on separate heads. The MPH component of this system proposed provides high-resolution imaging of activity in the structures of interest: the putamen, caudate and substantia nigra (SN), which lie in the central interior portion of the brain. In many multi-pinhole (MPH) SPECT imaging systems, improvement in image quality has been demonstrated with accurate modeling of the system point spread function (PSF) within reconstruction. To facilitate faster reconstruction and smaller memory overhead, simple PSF models are preferred such as a Gaussian. In many geometries a Gaussian model is reasonably accurate for the PSF. However, for different pinhole aperture shapes such as square or rectangular, and angled pinhole aperture planes, the measured PSF including all physical effects such as penetration and depth of interaction can have substantial non-Gaussian components. In this work, we illustrate with examples from our MPH collimator how stepwise regression analysis can be used to develop a generalized but parsimonious model for complex PSFs without a-priori knowledge of the exact analytical form of the PSF.