A new algorithm to harmonize SUV values obtained by different PET/CT scanners

389 Aim: Last generation PET/CT scanners implemented point-spread function (PSF) correction and time-of-flight (TOF) data in reconstruction software. These novelties cause an increase of Standardized uptake values (SUV) compared to conventional OSEM reconstruction, that might be advantageous for detection of small lesions, but could lead to errors in clinical scenarios in which SUVs are used for prognosis or response to therapy assessment. In this work we investigated a methodology to harmonize SUV values obtained by two PET/CT scanners of different manufacturer and generation. Materials and Methods: Phantom and patient data were acquired by a new PET/CT system equipped with LSO crystals, PSF and TOF algorithms and by a 10-years old PET/CT system equipped with BGO crystals (SIEMENS BIOGRAPh mCT and GENERAL ELECTRIC Discovery STE respectively). SIEMENS workstation (syngo®.via) is provided with a new quantification technology (EQ-Filter) that allows to harmonize SUVs, even if acquired on different scanners, without requiring modifications of the protocol or additional reconstructions. For phantom measurements we used a NEMA PET IQ phantom equipped with six spheres with internal diameter ranging from 10 to 37 mm. Two sets of 18F-FDG activity concentrations (on both spheres and phantom background) were tested obtaining lesion to background ratios of 8:1 and 4:1 respectively. Several different acquisition and reconstruction protocols were used in order to provide a general methodology to harmonize SUVs (independent of the specific protocol adopted in each institution). Data obtained on phantom measurements were validated on 7 oncologic patients who accepted to perform an extra acquisition (one bed position) on a different scanner. The evaluation regarded 39 small lesions (diameters ranging between 0.3 cm to 2.6 cm). Results: The main benefit of PSF+TOF systems is the increased percentage contrast of small lesions. In fact, the standard protocol used in our institution produced images with high resolution (3.7mm FWHM) and high contrast (the partial volume effect of the smallest sphere of 1cm was almost annulled). On the other hand, the curves of SUVmax values and of the percentage contrast on all spheres (measured according to NEMA standards) exceeded those obtained by the BGO PET/CT. Percentage contrast obtained in our institution, also applying the supplemental Siemens EQ filter, are shown in figure 1, together with the EARL accreditation limits on PET/CT scanners published by EANM in 2015. The discrepancies, without the smoothing filter, were as high as 140%, while they drop to 1-5% applying the optimized value of EQ filter (8mm in our case). For each reconstruction setting and for each combination of tomographs it is necessary to identify the optimal value of the EQ filter in order to minimize these SUV discrepancies. Data obtained on patients (on 39 lesions) using different EQ filter settings, confirmed the results of phantom measurements (figure 2), with Wilcoxon p-values