Energy-Based Scatter Estimation in Patient Scans Acquired with an SiPM PET Scanner

Abstract

Energy-based scatter estimation (EBS) is an alternative to single-scatter simulation (SSS) for PET image reconstruction in SiPM PET scanners. This paper describes clinical tests. List-mode data from two Biograph Vision scans were used: a whole-body <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> F-fluorodeoxyglucose oncology study (WB FDG); and a cardiac <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">82</sup> Rb blood flow study (RB) in one bed position. WB data were acquired 1 hour p.i. RB data were acquired from the start of infusion. Scatter estimation used 2D energy histograms in coarsely sampled sinogram bins, and was based on fitting these histograms to nine basis functions. In the case of RB, it was necessary to also model the background of prompt-gamma radiation in the 2D histograms, requiring sixteen basis functions. Two images were reconstructed in each case: one with EBS for scatter estimation and one with SSS. TOF-PSF reconstructions were used. The differences between EBS and SSS were quantified by comparing mean or maximum standardized uptake values (SUV) in various regions including a lung nodule. In WB, SUV derived with EBS were on average within 1.3% of SSS. Although SUV were only 2.8% elevated in the cerebrum, an unexplained 25% difference in scatter was noted in this area. Image values in the RB case differed by as much as 25% when EBS was used in place of SSS. In WB imaging, EBS yielded comparable images, was parameter-free and did not require CT images for scatter estimation.