Multicentre analysis of PET SUV using vendor-neutral software: the Japanese Harmonization Technology (J-Hart) study

BACKGROUND: Recent developments in hardware and software for PET technologies have resulted in wide variations in basic performance. Multicentre studies require a standard imaging protocol and SUV harmonization to reduce inter- and intra-scanner variability in the SUV. The Japanese standardised uptake value (SUV) Harmonization Technology (J-Hart) study aimed to determine the applicability of vendor-neutral software on the SUV derived from positron emission tomography (PET) images. The effects of SUV harmonization were evaluated based on the reproducibility of several scanners and the repeatability of an individual scanner. Images were acquired from 12 PET scanners at nine institutions. PET images were acquired over a period of 30 min from a National Electrical Manufacturers Association (NEMA) International Electrotechnical Commission (IEC) body phantom containing six spheres of different diameters and an (18)F solution with a background activity of 2.65 kBq/mL and a sphere-to-background ratio of 4. The images were reconstructed to determine parameters for harmonization and to evaluate reproducibility. PET images with 2-min acquisition × 15 contiguous frames were reconstructed to evaluate repeatability. Various Gaussian filters (GFs) with full-width at half maximum (FWHM) values ranging from 1 to 15 mm in 1-mm increments were also applied using vendor-neutral software. The SUV(max) of spheres was compared with the reference range proposed by the Japanese Society of Nuclear Medicine (JSNM) and the digital reference object (DRO) of the NEMA phantom. The coefficient of variation (CV) of the SUV(max) determined using 12 PET scanners (CV(repro)) was measured to evaluate reproducibility. The CV of the SUV(max) determined from 15 frames (CV(repeat)) per PET scanner was measured to determine repeatability. RESULTS: Three PET scanners did not require an additional GF for harmonization, whereas the other nine required additional FWHM values of GF ranging from 5 to 9 mm. The pre- and post-harmonization CV(repro) of six spheres were (means ± SD) 9.45% ± 4.69% (range, 3.83–15.3%) and 6.05% ± 3.61% (range, 2.30–10.7%), respectively. Harmonization significantly improved reproducibility of PET SUV(max) (P = 0.0055). The pre- and post-harmonization CV(repeat) of nine scanners were (means ± SD) 6.59% ± 1.29% (range, 5.00–8.98%) and 4.88% ± 1.64% (range, 2.65–6.72%), respectively. Harmonization also significantly improved the repeatability of PET SUV(max) (P < 0.0001). CONCLUSIONS: Harmonizing SUV using vendor-neutral software produced SUV(max) for 12 scanners that fell within the JSNM reference range of a NEMA body phantom and improved SUV(max) reproducibility and repeatability.