Effects of motion and tissue weighting on PET myocardial blood flow estimates

Objectives The objectives of this study was to identify prevalence of respiratory motion in dynamic PET myocardial perfusion images, its effects on myocardial blood flow (MBF) and coronary flow reserve (CFR) estimates and to investigate the implications of the weighting model deployed during kinetic fit to minimize effects of motion.

Methods This study used 53 randomly selected patients referred for dynamic stress/rest Rb-82 PET imaging. Dynamic image series were iteratively reconstructed. The left ventricular (LV) blood pool and tissue time activity curves (TAC) were sampled using ROI and fit to a compartment model with uniform or tissue weighting for MBF and CFR quantification using Corridor4DM software (INVIA). Tissue weighting incorporated the frame duration, decay correction factor, and activity concentration. Motion was assessed visually from the dynamic series and categorized with having minimal, moderate, or severe motion. Studies with moderate or severe motion were manually corrected for comparison.

Results Moderate and severe motion was observed in 46% and 39% of the studies respectively. Motion was most prominent in the blood pool phases with 55% of the motion in the craniocaudal direction. Compared to the motion corrected images, images with moderate motion had global stress MBF values 3.2% higher when uniform weighting was used. Using tissue weighting only reduced those differences by 0.5%. Studies with severe motion had greater stress MBF differences: 5.8% (uniform) and 5.4% (tissue). Regionally stress MBF values in the RCA varied the greatest: 20.2% (uniform) and 19.9% (tissue).

Conclusions Respiratory motion is problematic in the blood pool phase due to fast kinetics of the tracer and the short time sampling. Increasing differences in MBF and CFR estimates were observed between studies with minimal, moderate, and severe motion. Tissue weighting has limited impact on reducing the increase in MBF and CFR estimates caused by motion suggesting motion correction is needed to reduce bias in flow estimates.