data-driven, projection-based respiratory motion compensation of pet data for cardiac pet/ct and pet/mr imaging

Respiratory patient motion causes blurring of the pet images that impact accurate quantification of perfusion and infarction extents in pet myocardial viability studies. in this study, we investigate the feasibility of correcting for respiratory motion directly in the pet-listmode data prior to image reconstruction using a data-driven, projection-based, respiratory motion compensation (dpr-moco) technique. the dpr-moco method was validated using simulations of a xcat phantom (biograph mmr pet/mr) as well as experimental phantom acquisitions (biograph mct pet/ct). seven patient studies following a dual-tracer (18f-fdg/13n-nh3) imaging-protocol using a pet/mr-system were also evaluated. the performance of the dpr-moco method was compared against reconstructions of the acquired data (no-moco), a reference gate method (gated) and an image-based moco method using the standard reconstruction-transform-average (rta-moco) approach. the target-to-background ratio (tbrlv) in the myocardium and the noise in the liver (covliver) were evaluated for all acquisitions. for all patients, the clinical effect of the dpr-moco was assessed based on the end-systolic (esv), the end-diastolic volumes (edv) and the left ventricular ejection fraction (ef) which were compared to functional values obtained from the cardiac mr. the dpr-moco and the no-moco images presented with similar noise-properties (cov) (p = .12), while the rta-moco and reference-gate images showed increased noise levels (p = .05). tbrlv values increased for the motion limited reconstructions when compared to the no-moco reconstructions (p > .05). dpr-moco results showed higher correlation with the functional values obtained from the cardiac mr than the no-moco results, though non-significant (p > .05). the projection-based dpr-moco method helps to improve pet image quality of the myocardium without the need for external devices for motion tracking.