Interaction of patient motion and RV spillover correction on septal wall MBF and MFR

Objectives: Contemporary clinical software tools have greatly improved the efficiency and robustness of routine myocardial blood flow (MBF) estimation by dynamic cardiac PET. However, such tools do not generally correct for patient motion, and are particularly vulnerable to quantitative errors during the first-pass blood pool phase of the dynamic sequence. Spillover from the right ventricle (RV) into the interventricular septum during this phase can also bias MBF estimates [Hove, 1998; 2003] [Hermansen 1998]. Because uncorrected motion may result in increased apparent spillover, we investigated the potential for RV spillover correction to mitigate the effects of patient motion on regional septal MBF and flow reserve (MFR).

Methods: Consecutive patients (N=105, 72 M; age 64±12 yrs) underwent clinically indicated rest/regadenoson stress myocardial perfusion PET using Rb-82 in list-mode. Dynamic data were binned into 30 frames (16x5s, 6x10s, 3x20s, 4x30s, 1x80s). Patient motion between dynamic PET frames was manually corrected by expert users using the myocardial contours determined from the last 5 minutes of dynamic PET data as a reference. The frequency and magnitude of motion artifact was determined from motion correction vectors. 1-tissue-compartment kinetic models were fit to dynamic data with blood pool spillover correction for both LV and RV (RV+) and with LV spillover correction only (RV-). Bland-Altman plots and relative differences in segmental MBF and MFR estimates in the septum were compared with and without correction for motion (MC+/MC-) and RV spillover (RV+/RV-) with respect to RV+MC+. Results are reported as median(IQR). RESULTS: The median number of frames requiring motion correction were 19 (5, 24) and 18 (2, 23) at rest and stress, respectively. During the blood pool phase (frames 1-20) the median frequency of motion correction was significantly higher than during the tissue uptake phase (frames 21-30) [0.61 vs 0.44 at rest, 0.59 vs 0.38 at stress, both p<0.001]. The magnitude of motion correction was also higher during the blood pool phase [1.3 (0, 4.7) mm vs 0.0 (0, 2.3) mm at rest, 1.3 (0, 5.5) mm vs 0.0 (0, 2.3) mm at stress, both p<0.001]. 525 septal segments were analyzed. Median differences in septal segmental MBF and MFR relative to RV+MC+ were < 10% in all cases (Figure). When motion was corrected (MC+), the effect of uncompensated RV spillover was moderate (Figure, left column, note smaller y-scale compared to center & right columns), with fewer than 10% of septal segments having > 10% MBF or MFR differences (Table). However, without motion correction (MC-) the variability was greatly increased (Figure, middle & right columns). In addition, the number of septal segments with relative differences > 10% were also significantly increased (Table). CONCLUSIONS: The incidence and magnitude of motion artifacts were greater in the blood pool phase. The predominance of motion artifact during this phase tended to amplify RV spillover bias. When motion correction was omitted, large increases in septal MBF and MFR variability were observed which were not mitigated by RV spillover correction. Correction for dynamic PET motion artifact is clinically necessary to reduce regional MBF and MFR variability. RV spillover correction provides an additional smaller reduction in variability but its clinical impact is yet to be determined.

• Download figure
• Open in new tab
• Download powerpoint