Abstract
486
Objectives Post-reconstruction correction of patient body motion during dynamic positron emission tomography (PET) can leave up to 50% error in the measured myocardial blood flow (MBF) values, that may contribute to misdiagnosis of coronary artery disease. The aim of this study was to evaluate the performance of an algorithm to correct for rigid patient body motion before image reconstruction.
Methods Patient-derived time-activity curves (TAC) for Rb-82 PET myocardial perfusion imaging were used in conjunction with a digital NCAT phantom and the analytical simulator ASIM to produce realistic projection data with noise for a 35 cm phantom. Instantaneous shifts of ±1 cm at 90, 120, 150, 180 and 240 s in the patient body location were simulated. Motion detection, CT alignment, and subsequent shift-correction was performed by minimization of the Helgason-Ludwig Consistency Conditions for the parallel-beam 2D Radon transform. Dynamic images were reconstructed using the OSEM-One Step Late iterative algorithm as implemented in the Software for Tomographic Image Reconstruction (STIR). Blood flow quantification was calculated with FlowQuant (UOHI) using the 1-tissue-compartment model of Lortie (2007), with both regional (Flow) and global (FlowRC) partial volume correction (PVC). A baseline without motion was used to calculate relative MBF error.
Results The digital phantoms were successfully motion-corrected to 0.2 voxels (0.6 mm) of the ideal position on average. Largest deviations in correction were on the order of 0.8 voxels (0.25 cm). Automated correction of simulated NCAT phantom motion resulted in a drop of the median relative MBF error from 10.1% to 1.8%.
Conclusions Projection-based motion correction dramatically reduced the body motion-induced error in MBF measurements for Rb-82 PET simulations. The proposed, automated motion correction method appears promising for correcting patient motion prior to reconstruction, warranting further investigation.