A Source-Traced PET/CT Scanner Performance Database (2004–2024): The Sphere-Size Dependence of Time-of-Flight Contrast Recovery

Background: Small-lesion quantification is a key task in diagnostic PET/CT. However, published performance data span two decades and multiple NEMA NU 2 revisions, complicating comparisons. We developed a source-traced database to examine how reconstruction-associated improvements in hot-sphere contrast recovery vary with sphere diameter, providing reference data for phantom-based protocol optimisation. Methods: Published NEMA and phantom measurements were source-verified, linked to scanner geometry, and annotated with acquisition and reconstruction conditions. Contrast recovery at a 4:1 sphere-to-background activity ratio (151 measurements; 13 scanners) was analysed using mixed-effects meta-regression with a scanner random intercept. Results: The database included 28 scanner models (2004–2024), 610 measurements and 43 publications. Descriptively, sensitivity was 3.8–10.9 cps/kBq for lutetium-based analogue scanners, 12.1–47.6 for conventional-geometry digital scanners and 176–192 for total-body scanners. Time of flight (TOF) was associated with a 7.9-percentage-point-higher contrast recovery (95% CI 4.8 to 11.1), but the association varied with sphere diameter (p < 0.001) and was significant only at 10–17 mm (11.4, 9.9 and 7.0 points); estimates at 28 and 37 mm rested on four to five scanners and were imprecise. The median between-laboratory difference under matched conditions was 5.7 points. Point-spread-function and regularised-reconstruction effects were not robust. Conclusions: In phantom measurements, TOF-associated gains were concentrated at 10–17 mm. These data provide reference values for phantom-based protocol optimisation, harmonisation and acceptance testing; they do not measure lesion detectability or diagnostic accuracy, which require patient-based validation.