Optimization of scan duration and image smoothing for obese patients in PET imaging

Objectives Recently we derived functions correlating PET image noise to scan duration (SD) and image smoothing (IS) for patient studies with different BMI (Med Phys, 2010; 37:3139P). These functions can be used to optimize PET SD and IS for a preset image noise. Our aim is to validate the derived functions using phantom studies and to find the optimal SD and IS that result in a lower image noise for different BMIs while maintaining ≥90% SUVmax calculated when using our clinical settings of SD/IS = 3min/6mm.

Methods An IEC phantom containing 6 spheres (1-3.7 cm) was scanned on a GE DVCT PET/CT scanner. The sphere to background ratio was 5.1:1. A uniform water phantom injected with 37 MBq was placed behind the IEC phantom to represent a scatter source. PET data was acquired in 3D for 20 min using LIST mode. The same acquisition was repeated while 8 saline bags (250ml each) were attached to the IEC phantom to increase its BMI from 27 (low) to 32 (high). The acquired PET data was rebinned and processed using different combinations of SD/IS for the low and high BMI phantom studies. To validate the previously derived functions, image noise was calculated from 300 randomly selected voxels in the IEC background and compared to that from the derived functions. Optimal SD and IS was then determined from the settings that result in lower image noise for each BMI while maintaining ≥90% SUVmax for each sphere compared to our clinical settings.

Results Noise from phantom scans was on average 9±5% different from that predicted by the derived functions. SD/IS that result in 90% and 95% SUVmax of the clinical settings for low (high) phantom BMI were 4/7.6 and 5/6.8 (8/8.3 and 12/6.9 min/mm) respectively.

Conclusions Our previously derived functions based on patient data were validated by this phantom study. A SD/IS of 4/7.6 and 8/8.3 (min/mm) can be used for patients with BMI of 25-30 and 30-35 respectively to obtain an image with lower noise than that derived when using our clinical settings while maintaining an SUVmax ≥90%