Respiratory-motion errors in quantitative myocardial perfusion with PET/CT

Respiratory motion is known to cause errors in whole-body oncologic and static cardiac imaging with PET/CT. These errors are caused by the difference in acquisition times of the PET and CT data sets leading to inconsistencies and hence artifacts when the CT scan is used for attenuation correction (CTAC). The purpose of this study was to use computer simulations to investigate how quantitative imaging of myocardial perfusion with dynamic Rb82 PET/CT may be affected by respiratory motion. The NCAT anthropomorphic computer phantom was used to generate uniform-activity images at each of 10 respiratory phases and 17 dynamic frames. PET projection data for each of these 170 images were generated using the SimSET Monte Carlo simulator. The GE Discovery LS PET/CT was modeled and 400 M photon histories were tracked for each simulation. Images were reconstructed using OSEM and 4 different approaches to CTAC: phase-matched CTAC, correction with a single-phase CT (end- inspiration, end-expiration, and mid-inspiration), correction with a CT averaged over the respiratory cycle, and correction with a CT that was a voxel-by-voxel maximum over the respiratory cycle. The dynamic image sets were then processed using software developed in-house for the kinetic analysis of myocardial perfusion data. Images of Kl (blood-flow) were converted to polar maps and compared point-by-point and by segment using 17-segment regional analysis. Comparing all results to those of the phase-matched correction, we found that a single-phase correction had mean segmental errors as high as 20% (end- inspiration) with mid-inspiration correction providing the least error at 6%. An average CTAC had errors as high as 12% in the mid-inferior wall. The max-CTAC approach produced errors as high as 21%. Also of note, though, was that the phase-matched polar map was not uniform, as expected, with a visible decrease in blood-flow in the inferior wall. This deficit is potentially caused by motion-blurring leading to i- nterference from the activity in the stomach and liver through the model fitting of the spill-over correction term. We conclude that respiratory motion can lead to errors in quantitative estimates of blood-flow obtained from dynamic Rb82 PET perfusion studies. A CT map acquired at the mid-respiratory phase provides an accurate CTAC correction, but is not practical to acquire. An average CTAC provides the most accurate, practical solution to the respiratory-motion problem, however, it still produced segmental errors as large as 12% in this simulation. Errors in the inferior wall of the heart were observed with all corrections and may be related to motion-blurring of activity from extra-cardiac organs into the heart. The source of the inferior-wall deficits requires further investigation.