Sinogram-based Alignment Estimation of PET Detector Blocks

Abstract

Mechanical inaccuracies in PET detector installation due to errors and imprecisions during manufacturing pose significant challenges to image quality. If unaccounted for gamma interaction positions and consequently resulting lines of response are mispositioned and lead to reduced signal-to-noise ratios in reconstructed images. We propose to optimize transaxial detector alignment parameters using stochastic gradient descent by matching a differentiable version of the sinogram with the analytical solution of the tracer line integral values. Our method conceptually allows for the usage of arbitrary tracer distributions, rendering the need for specialized phantoms and acquisition protocols obsolete, and paving the way towards the usage of clinically available phantoms for fast and regular detector alignment calibration. We utilize the Adam optimizer with an adaptive normalization strategy and convex hull regularization to incorporate prior knowledge on established mounting precisions. To deal with known phantoms placed in unknown positions, we incorporate a clustering procedure in the noisy reconstructed images to adaptively compute target tracer distribution and line integral values. Our study involved GATE simulations of multiple point sources with known and unknown positions inside a small animal PET scanner. The detector configuration was subsequently subjected to simulated alignment noise. Our method is able to estimate positions with ~300μm and axes with ~0.7° precision, even in cases where point source positions were unknown. Our research demonstrates that the fitting of sinograms proves a valid strategy for transaxial detector alignment with the potential for extension towards z-axis alignment.

Authors 2

1. Julian Thull Aachen Institute for Experimental Molecular Imaging Department of Physics of Molecular Imaging Systems

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Institute for Experimental Molecular Imaging,Department of Physics of Molecular Imaging Systems,Aachen,North Rhine-Westphalia,Germany

2. Volkmar Schulz Aachen Institute for Experimental Molecular Imaging Department of Physics of Molecular Imaging Systems

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University,Institute for Experimental Molecular Imaging,Department of Physics of Molecular Imaging Systems,Aachen,North Rhine-Westphalia,Germany