Spatial variability in a very high sensitivity PET system

Objectives A new PET/CT system with extremely high sensitivity (22 cps/kBq - NEMA NU2) is available (Discovery IQ - 5 ring, GE Healthcare Technologies). This sensitivity is due to the use of 3 cm-long bismuth germanate detectors configured in a cylinder of 74 cm diameter and 26 cm length. Non-negligible gaps between the 5 cm x 5 cm detector blocks in the circumferential direction lead to more variability (spatially) in the sensitivity than systems without such gaps. This is an investigation of that variability and its significance, comparing with a lower sensitivity PET/CT with negligible gaps between detectors.

Methods Measurements were performed on a Discovery IQ (5 ring) and a Discovery 690 (also GE Healthcare). The spatial variability of the sensitivity was measured with a 16 cm Ge-68 line source. The line source was placed on the PET central axis (R=0). The source was also centered axially. The source was then stepped increasingly off-center (while remaining parallel with the scanner axis) for a series of scans in step sizes 2 mm (from R=0 to R=1 cm), 5 mm (from R=1 cm to R=7cm, and 10 mm (from R=7cm to R=20cm) . A 10 s acquisition was performed at each position. Random-corrected count rates were calculated for each position for a relative measure of sensitivity at each R. To compare sensitivity with less position dependence, a 32 cm long, 10 cm diameter phantom filled with F-18 solution scanned on both systems. The diameter was chosen to be large enough to average over some of the spatially-varying sensitivity effects while also producing fewer scattered photons than a larger phantom would yield.

Results The DIQ sensitivity profile (vs R) was sharply peaked at the tomograph center, dropping to 94% of that peak at R=4 mm and to a minimum of 75% at R=2 cm, confirming that the R=0 NEMA sensitivity measurement must be performed with great care in positioning the line source. The variation continued with a ~3.4 cm period for increasing R, but was much less intense than at the center. There was also an overall gradual increase in sensitivity with R. The D690 sensitivity had no significant local variations and only the gradual increase with R. The 10 cm cylinder yielded a sensitivity 2.48X higher on the DIQ than the D690, whereas NEMA line source measurements (reflecting the central sweet spot) are 2.75X higher.

Conclusions While all PET tomographs have a line source sensitivity that varies with R, the Discovery IQ sensitivity has a more local and dramatic effect. NEMA sensitivity measurements for acceptance or QA, and any routine measures of sensitivity based on a central line must be performed carefully, since the error due to mispositioning the source by a few mm may be greater than typically sensitivity variations in systems. A cylindrical phantom would be a better choice for routine QA and, perhaps, for measuring the true sensitivity improvement of this system over other tomographs.