Abstract
568
Objectives: Recent hardware and reconstruction advances in commercial PET/CT scanners have resulted in substantial gains in signal-to-noise ratio and reconstructed spatial resolution. Currently available commercial phantoms are highly symmetrical and contain idealized resolution features that poorly predict how these newer systems perform clinically. This report describes the design, construction and testing of an anthropomorphic oncology phantom designed to quantitatively characterize PET/CT scanner performance across a meaningful range of object sizes using clinically relevant conditions.
Methods: The SNMMI Clinical Trials Network (CTN) oncology torso phantom was redesigned and now includes: 1) a simulated low-density lung field, 2) A set of six spheres with internal diameters ranging from 10-37 mm, asymmetrically placed inside the torso allowing measurement of contrast recovery coefficients (CRC) comparable with NEMA Image Quality (IQ) phantom measurements, 3) a 7mm sphere to assess the detectability of small lesions, 4) two 10mm spheres separated by 1mm to characterize lesion separation. The spheres and internal support structures are 3D printed to assure rigid, reproducible positioning of the lesions within the phantom volume. A total of 12 spherical lesions ranging from 7 to 37mm are present. The precisely consistent geometry enables a fully-automated analysis of standardized uptake value metrics (max, peak, mean) for the spheres, as well as background measurements to assess calibration accuracy. Fully-automated analysis software has been developed to detect and segment the spheres using the known geometry to standardize results assessment. The sphere detection algorithm has been demonstrated to be robust to phantom orientation (Ulrich, Med Phys, 2017). Two different 3D printing plastics were used to optimize for the relevant feature: printing resolution (sphere walls) and ruggedness (support structures).
Results: Attenuation properties of the plastics were measured to assure reasonable compatibility with PET/CT attenuation lookup tables. The sphere plastic attenuation (HU=145) is sufficiently different from water that PET/CT fusion accuracy can easily be assessed in all three dimensions as part of the scanner assessment. 3D printed plastic for support structures had HU = -55. Three identical prototype phantoms were constructed and assessed in conjunction with three PET/CT scanners of different vintage and technology: A non-TOF PET/CT, a current generation photomultiplier tube (PMT) TOF PET/CT, and a Silicon Photomultiplier (SiPM) array state-of-the-art TOF PET/CT. Where available, point response function and regularized reconstructions were enabled to compare with standard reconstructions. CRC curves were generated from the 10 to 37mm spheres in the CTN phantom. These data compared favorably with corresponding CRC data obtained from a conventional NEMA IQ phantom acquired under similar conditions for all scanner systems. The 7mm sphere was not visualized at 4:1 contrast with the non-TOF PET/CT. It was comparable to noise texture with the PMT TOF PET/CT, although arguably detectable, and clearly visualized with the SiPM TOF PET/CT. Advanced reconstructions enhanced detectability in the two TOF scanners. Advanced reconstructions were not available on the non-TOF PET/CT. Conclusions: A new 3D printed anthropomorphic PET oncology phantom has been designed to: generate NEMA IQ CRC curves that are comparable to existing NEMA methodology; assess lesion detectability in a clinically relevant way; measure scanner calibration accuracy, and determine PET/CT alignment. The proposed phantom is compatible with use for scanner validation for clinical trials and to fulfill several of the new Joint Commission phantom-based quality control requirements. It is intended that the modified CTN phantom described above will be commercially available in the near future. Funding: R01CA169072