Standardized quantitative characterization of PET/CT scanners

588 Objectives Current emphasis on achieving accurate quantitation of PET/CT systems is occurring at the same time that performance is being enhanced through time-of-flight technology and increasingly sophisticated image reconstruction algorithms. This work reports on a phantom imaging and analysis protocol allowing for quantitative characterization of current model PET/CT scanners by using modifications of the NEMA IEC body phantom with the ultimate goal of identifying harmonized reconstructions for clinical trials. Methods To more fully characterize the recovery coefficient curve space beyond the standard 10-37mm diameter spheres, 7 9half-step9 spheres (8.5, 11.5, 15, 19.5, 25, 32.5, 44mm) and ellipsoids compatible with the NEMA IEC phantom were designed in AutoCADTM and converted to .STL files for use with rapid prototyping (stereolithography) manufacturing. The prototype spheres were made from clear polycarbonate-like material with tolerances of 0.25mm. The spheres were distributed to 4 imaging sites with detailed fill and imaging procedures. The two-day imaging procedure imaged the standard sphere set (all hot) on day one and the half-step sphere set on day two at 9.7:1 contrast consistent with EARL protocols. An additional 4.85:1 acquisition for lower contrast characterization was also performed. Results Semi-automated analysis techniques to generate contrast recovery coefficients for SUV max, peak, mean, and threshold were developed, tested, and used for consistent analysis. Contrast recovery curves were generated for multiple systems and reconstructions with sphere sizes ranging from 8.5-44mm. Results suggest quantitatively harmonizing reconstructions will be challenging due to differing reconstruction approaches by the vendors. Conclusions A simple, reproducible method for consistent quantitative characterization of PET/CT systems was developed and validated at multiple sites and scanners, which will be used to determine the approach for harmonization. Research Support Research reported in this abstract was supported by the National Cancer Institute of the National Institutes of Health under Award Number R01CA169072