Abstract
1664
Objectives: PET images are routinely reconstructed using iterative techniques which are statistical and non-linear in nature. The aim of this abstract is to investigate the effects of scan duration and count density on the quantitative accuracy of PET images.
Methods: An IEC phantom containing 6 spheres (1-3.7 cm) was scanned on a GE DRX PET/CT scanner. SBR was 7:1. PET data was acquired in 2D and 3D for 3 minutes using LIST mode and then rebinned into 11 different scan durations ranging from 5 sec to 3 min. The same experiment was repeated three times at one hour intervals (4 total scans) to assess the impact of count density on quantitative accuracy. All PET images were reconstructed using OSEM (2 iterations 21 subsets). ROIs were drawn on all 6 spheres of the resulting images. Mean and max activity concentration (AC) as well as CNR was calculated and plotted for all spheres in all scans. In addition, SNR was calculated as the ratio of mean sphere AC to the STD of 60 randomly selected pixels in the background.
Results: Max AC increased as scan duration and count density decreased for all spheres in both 2D and 3D images. This increase exceeded 10% for scan durations less than 30-40 sec. Mean AC was relatively uniform in all images. Noise decreased faster with scan duration in 3D compared to 2D and SNR reached a plateau faster in 3D than 2D. Moreover, noise increased while SNR and CNR decreased with decreasing count density.
Conclusions: Decreasing scan duration can have adverse effects on the quantitative accuracy of PET AC. These results might impact outcome measures particularly in dynamic scans where scan duration are initially very short.
- Society of Nuclear Medicine, Inc.