Harmonization of PET Image Reconstruction Parameters on Multiple PET/MRI System

Richard Laforest; Mohammad Khalighi; Darrin Byrd; An Hongyu; Peder Larson; John Sunderland; Paul Kinahan; Thomas Hope

Objectives: Recently simultaneous PET/MRI cameras a have been made available commercially by two manufacturers. These two cameras vary in their quantitative PET performance due to inherent differences in their construction and to the varied image reconstruction settings employed. The resulting variability in quantitative accuracy confounds the ability to meaningfully combine and compare data between the two scanners, and especially in clinical trials that aim at including data from cameras from both manufacturers. In this work, we aim at defining optimal harmonized image reconstruction parameters that lead to most comparable contrast recovery curves across simultaneous PET/MRI scanner models. Methods: The NEMA NU-2 image quality phantom was imaged on the GE Signa and on the Siemens mMR PET/MRI scanners. The phantom was imaged with its standard sphere set (diameter 10, 13, 17, 22, 28, 37mm) and with a custom spheres set (diameter: 8.5, 11.5, 15, 25, 32.5, 44 mm) filled with a ratio of sphere to background of 9.7:1 using a standardized methodology. The background activity was set at approximately 1700 Bq/mL and data were acquired for 30 min. Images were reconstructed with the manufacturer provided iterative image reconstruction algorithms with and without point spread function (PSF) compensation. On the GE Signa, images were reconstructed with 2 and 4 iterations, 16 subsets, using Time of Flight (TOF). On the Siemens mMR, images were reconstructed with 1 to 4 iterations, 21 subsets. For both scanners, a post reconstruction Gaussian filter of 3 to 7 mm in steps of 1 mm were applied to generate 20 different reconstructions on the GE Signa and 40 different image reconstructions on the Siemens mMR. Attenuation correction was provided from a scaled Computed Tomography (CT) image of the phantom registered to the DIXON MR attenuation images, and included attenuation from the scanner couch. For each of these clinically relevant image reconstruction parameter sets Contrast Recovery Coefficient (CRC) curves were determined for the SUVmean, SUVmax and SUVpeak within each sphere. The mean squared differences of CRC over sphere size was computed and used to rank the similarity of image reconstruction combination pairs for the two scanners. The image reconstruction parameter set with the lowest mean squared difference was identified as the best candidate reconstruction for each vendor for harmonized PET image reconstruction. Results: The range of clinically relevant image reconstruction parameters demonstrated widely different quantitative performance across the two manufacturers. By varying both the number of iterations and post reconstruction filter level, bands of CRC curves were obtained that showed significant overlap between the two PET/MR manufacturers. The best match of CRC curves are obtained with: 2 iterations -7mm Filter with PSF on the GE Signa and 4 iterations, 5mm filter on the Siemens mMR for the SUVmean, 6 iterations -6mm filter on the GE Signa and 4 iterations, 5mm filter on the Siemens mMR for the SUVmax and 4 iterations -4mm Filter on either scanner for the SUVpeak. Over all reconstructions, the mean squared differences between CRCs for the scanners were 6.7% and 15.4% for mean and max recovery, respectively. These were reduced to less than 2% for optimally matched reconstruction settings. Conclusions: For two commercially-available PET/MRI scanners, user-selectable parameters that control iterative updates, image smoothing, and PSF-modeling provide a range of contrast recovery curves that allow harmonization. This work demonstrates that essentially identical CRC curves can be obtained on the commercially available scanner by a proper choice of image reconstruction parameters.

Harmonization of PET Image Reconstruction Parameters on Multiple PET/MRI System

View Online

Abstract

Details

Metrics