Abstract
2919
Introduction: In a typical brain PET/MR scan, the PET acquisition is done during the entire exam, which may last between 30-60 minutes. While the subject's head is restrained in the head coil, some subjects, especially patients with a movement disorder or Alzheimer's disease, exhibit some degree of motion during the PET acquisition which may result in image blurring, quantitative errors due to mismatched attenuation correction, and mis-registered PET and MR images. In this study, we have analyzed 220 PET/MR brain scans in our center and estimated the head motion directly from the PET data [1], to investigate the extent of the subject motion during the scan.
Methods: 220 PET/MR scans from multiple studies were selected and the head motion was measured using a motion estimation technique based on the PET data using image registration of very short frames [1]. This included 66 scans of 18F PI2620 from a dementia study, 53 scans of 18F -FDG from a head trauma and Glioblastoma study, 44 scans of 18F-Florbetaben (FBB) from a dementia and Parkinson's disease study, 20 scans of 18F-DASA from a pyruvate kinase M2 evaluation study, 14 scans of 18F -FTC from a pain study, 14 scans of 18F-FMZ from a Fragile-X study and 9 scans of 11C -Raclopride from a mild depression study. All subjects were consented per local IRB guidelines before enrollment.
For each study, two points, one 70 mm anterior and one 70 mm posterior to the approximate center of the brain, were moved according to the estimated motion and their absolute displacement from the reference position was measured. The maximum displacement of the anterior and posterior points was recorded, and their effective displacements were calculated using a root-mean-square measure.
Also, the anterior and posterior point displacements were used to determine the motion group. The median displacement over time was calculated for each point, and the motion group is determined based on the point with the larger median displacement: ‘Low': < 1 mm, ‘Medium': > 1 mm and < 2 mm and ‘High': > 2 mm. Then for each subject, the maximum displacement of anterior and posterior points was rounded to the nearest mm and the percentage of total scan time that subject spent on each value (1 - 60 mm) was calculated. We are assuming that pure rotations about the center anterior-posterior axis of the brain are rare. The average scan-time percentage for each displacement was calculated over all scans.
Results: Figure 1 shows a typical motion measurement during a 1-hour brain exam of a subject injected with 8 mCi of FDG. The rotation and translation parameters of the brain motion are plotted as well as the absolute displacement of two points 70 mm away from the center of the brain in anterior and posterior directions. Figure 2 shows the maximum and effective displacement of the anterior and posterior points for 220 exams. The average scan time for the exams is 48 ± 14 min as shown in figure 2. The average effective displacement is 3.6 ± 4.5 mm and 2.8 ± 3.1 mm for anterior and posterior points and their maximum displacement is 8.2 ± 10.2 mm and 5.7 ± 6.3 mm, respectively. Figure 3 shows the motion group (i.e., ‘Low', ‘Medium' or ‘High' motion) for all scans. About 47% of subjects show high motion and were not able to hold still during their exam. Figure 4 shows the percent of the total scan-time that subjects have spent on average around each displacement value (0 – 25 mm) for anterior and posterior points. It shows that on average, subjects have moved less than 0.5 mm in ~20% of the scan time have moved more than 1.5 mm in ~60% of the scan time.
Conclusions: The analysis of 220 PET/MR brain scans shows that motion is prevalent and 47% of our subjects showed high motion in their exam and it is difficult for subjects to hold still for the entire PET scan time. In addition to the image blurring artifact, motion may cause quantitative errors due to mismatched attenuation correction and motion correction methods such as [1] or [2] will help to address these problems.