Feasibility of a brain PET harmonization program for state of the art PET/CT systems

Ronald Boellaard, Sandeep S. V. Golla, A Kaleep, Maqsood Yaqub, Terez Sera, Sjoerd Rijnsdorp, Rosalie Vered Kogan, Klaus L. Leenders, Adriaan A. Lammertsma

Research output: Contribution to journalMeeting AbstractAcademic


Purpose/Introduction: Use of brain PET studies in multicentre trials or as a quantitative imaging biomarker for (automated) differential diagnosis of neurogenerative diseases require harmonized quantitative image characteristics. In this study we explored the feasibility of developing a harmonizing performance standard for brain PET studies on state of the art PET/CT systems. Subjects & Methods: In this exploratory study 6 state of the art PET/CT systems were included: Philips Gemini TF, Ingenuity TF and digital Vereos systems, 2 Siemens Biograph mCTs and a GE 710. Only systems with EARL compliant (calibration and image quality) performances were included. A 30 min dynamic PET scan of the 3D Hoffmann brain phantom was acquired. The phantom was filled with an exact known FDG stock solution (aimed at 40 kBq/mL). Each scan was reconstructedusing various clinically relevant reconstruction settings. Depending on PET/CT system reconstruction settings were varied as follows: time of flight (TOF) on/off; resolution modelling (RM) on/off, voxel size, number of iterations/subsets and Gaussian smoothing FWHM (mm). The reconstructed images were analysed using a coregistered eroded binary map of both grey (GM) and white matter (WM). GM and WM recovery coefficients were calculated as the ratio of observed and expected activity concentrations. Results: For all systems distinct differences in both GM and WM recoveries and GM/WM ratios were observed between reconstructions that did or did not apply RM. Across the various systems/reconstructions a harmonized GM recovery between 0.77 and 0.85 (RM OFF) or between 0.81 and 0.94 (RM ON) seems feasible. WM recoveries (0.25 expected) were less affected by reconstruction settings, but showed a larger difference between Philips (0.28 to 0.33) versus Siemens (0.20 to 0.20) and GE (0.22 to 0.23) systems. GM/WM ratios were 4.2 to 4.4 for the Siemens and 3.7 to 4.0 for the GE systems, while the Philips systems showed somewhat lower values of 3.1 to 3.5 mainly because of difference in WM recovery. Discussion/Conclusion: Harmonization of PET/CT system performance for brain studies appears to be feasible, in particular for GM uptake assessment. Use of RM increases GM recovery at the cost of a wider (worse) harmonized performance range. There seems to be a vendor specific difference in WM recovery. The cause of this finding (possibly scatter correction) as well as its implication for PET/CT performance harmonization needs to be further explored. Currently, more data are being collected prospectively as part of JPND granted European networks.
Original languageEnglish
Pages (from-to)S224
Number of pages1
JournalEuropean Journal of Nuclear Medicine and Molecular Imaging
Publication statusPublished - 2017
Event30th Annual Congress of the European Association of Nuclear Medicine (EANM) - Vienna, Austria
Duration: 21-Oct-201725-Oct-2017
Conference number: 30

Cite this