Determining the impact of body truncation correction on PET SUV values using F-18 DCFPyL in Whole-Body PET/MR imaging

Objectives: Truncation techniques for attenuation correction maps (µ maps) have been widely used to better estimate SUV values for PET/MR imaging. Attenuation correction with truncation techniques can either use the PET emission data to create a maximum likelihood of activity and attenuation (MLAA) image or use the data acquired from a MR sequence that optimizes the readout gradients and compensates for artifacts due to B₀ inhomogeneities, HUGE (B₀ Homogenization Using Gradient Enhancement) [1]. Currently, there is a lack of data comparing the SUV values of reference tissues (parotid, blood pool and liver) and lesions for whole Body PET/MR imaging reconstructed with and without truncation techniques using a PSMA agent. This study aims to determine the variability in SUV values in reference tissues and lesions in whole body PET/MR imaging using F-18 DCFPyL reconstructed with (extended FOV using MLAA and HUGE) and without (2-point DIXON only) truncation techniques.

Methods: Whole body PET/MR images from 10 male patients injected with F18-DCFPyL (300MBq ±10%) were analyzed retrospectively. At 158mins (±10mins) post injection, images were acquired from vertex to mid-thigh using 5 bed positions at 2mins each. PET/MR data were reconstructed using the 2-point Dixon MR AC, HDPET (Point-spread function), 3 iterations with 21 subsets, 172x172 matrix, Gaussian filter with a FWHM of 5mm and absolute scatter correction. The data were also post processed with 3 different MRAC µ maps: 1) the standard (2-point DIXON), 2) the standard with the MLAA truncation technique and 3) the standard with HUGE truncation technique. Sphere VOIs were drawn over the right parotid gland (1cm³ ±0.5cm³), the left ventricle cavity as a representative of blood pool (3cm³ ±0.5cm³), the caudal aspect of the right liver lobe (3cm³ ±0.5cm³) and PSMA avid lesions on all 3 sets of images. SUV_max, and SUV_mean were calculated for all VOIs.

Results: Compared to PET AC images using the standard 2-point DIXON µ map, the PET data with µ maps using the HUGE and MLAA techniques have a higher increase in SUV_max and SUV_mean for the blood pool and liver than the parotid. The average increase using the HUGE technique compared to the standard for the blood pool was 16-20% while the increase was 4-5% using MLAA. As for the liver, the average increase using HUGE was 24-30% and 5-6% using MLAA. With the parotid, there was a change of less than 1% with HUGE and 2-3% with MLAA. The total average difference in SUV_max and SUV_mean of PSMA lesions increased by 35-40% using HUGE and 4-5% using MLAA. All results produced a p-value of less than 0.05.

Conclusions: When using truncation techniques for MRAC µ maps, there is an overall increase in SUV values for liver and blood pool reference tissues and lesions. Compared to the MLAA technique, the HUGE technique produced a higher percentage of change. While performing longitudinal studies using F-18 DCFPyL in whole body PET/MR imaging with semi-quantitative methods, the same truncation techniques should be maintained in order to preserve consistency.