Click here to


Are you sure ?

Yes, do it No, cancel

Comparison of An Anthropomorphic Phantom and the NEMA IEC Body Phantom for Harmonization of PET/CT and PET/MR Scanners

A Weisman1*, R Jeraj1,2, T Bradshaw1, (1) University of Wisconsin-Madison, Madison, WI, (2) University of Ljubljana, Ljubljana, Slovenia


(Sunday, 7/29/2018) 3:30 PM - 4:00 PM

Room: Exhibit Hall | Forum 9

Purpose: PET scanner harmonization is commonly performed by reconstructing phantom images with settings that minimize differences of recovery coefficients (RCs) across scanners. The goal of this study was to assess the impact of phantom complexity on harmonization results.

Methods: The phantoms used were the NEMA IEC PET phantom (six spherical inserts) and the RSD anthropomorphic thorax PET phantom (eight custom made shell-less spheres). Phantoms were scanned on a GE Discovery 710 PET/CT (D710), GE Signa PET/MR, and GE Discovery VCT PET/CT, matching counting statistics across scanners. PET images from the D710 and Signa were harmonized to the VCT by reconstructing with varied OSEM iterations, subsets, and post-filter widths, while matching PSF modeling, TOF status, and voxel dimensions to the VCT. Harmonization scores for each phantom and set of reconstruction settings were calculated as squared differences in RCs (maximum uptake normalized by expected uptake) from the VCT image, summed across ROIs. Resulting harmonization scores were compared between phantoms using Pearson’s correlation coefficient.

Results: Eighty reconstruction settings were assessed on the Signa and 43 on the D710. Correlation of harmonization scores between the NEMA and the RSD phantoms was strong for the Signa (R=0.92) and moderate for the D710 (R=0.74). Although optimal harmonization reconstructions settings were not identical between phantoms on either scanner, average differences in NEMA phantom RCs were low when using the optimal RSD settings (7.0%), and for the RSD phantom when using the optimal NEMA settings (3.7%) on the Signa. Differences were slightly greater on the D710, with an average difference in NEMA phantom RCs of 2.6% using the optimal RSD settings, and 11.5% vice versa. For comparison, average RC differences sometimes exceeded 40% before harmonization.

Conclusion: Differences between harmonization phantoms were insignificant on the PET/MR, but were more substantial on the PET/CT. Testing of more scanners is needed.


PET, Quantitative Imaging, Phantoms


IM- PET : Quantitative imaging/analysis

Contact Email