Multi-center comparison of a PET/CT calibration phantom for imaging trials

250 Objectives: We present the results of imaging a long-lived PET calibration phantom at multiple centers to assess how quantitative comparisons between the different scanners are affected by lesion size, scanner type, and local procedures. The goal is to determine the accuracy and precision due to instrumental factors for patient tracer uptake measurements in multi-center trials. Methods: The phantom, based on the NEMA NU-2 IQ phantom with Ge-68 (half life = 270 d) in epoxy, is designed to assess global accuracy, partial volume loss, reproducibility, and variations between scanners, protocols, and data reporting. The target:background ratio (T/B) for the 6 spheres was set to 4:1. Mean and max absolute activity concentration, T/B ratios, and SUV vs sphere diameter were measured. The phantom was imaged at 8 PET centers on 10 scanners manufactured by Siemens, Philips, and General Electric. Results: Average background SUV was 1.04±0.1. The recovery coefficient (RC) versus diameter depended on scanner type, image smoothing, whether absolute or relative mean or max values were reported. For typical clinical imaging protocols the coefficient of variation (COV) in RC across all scanners (averaged for all sphere diameters) was 8% if mean ROI values were used and 11% if max ROI values were used. Stochastic effects lead to a COV of approximately 3%, the residual COV due to deterministic differences between scanners and processing methods. Conclusions: The calibration phantom allows for direct comparison of quantitative results from sites in multi-center imaging trials using different scanners and/or different processing methods.