Use of a Qualification Phantom for Brain PET in a Multi-Center Consortium: A Collaboration Between the Pediatric Brain Tumor Consortium (PBTC) and the SNMMI Clinical Trials Network (CTN)

Objectives: To evaluate the image quality and quantitative capability for PET brain imaging across a multi-center consortium. METHODS: A phantom developed by the SNMMI CTN for the validation of clinical sites acquiring brain PET studies associated with clinical trials was sent to PBTC sites with PET capability. The phantom is comprised of 2 separate, fillable sections: one evaluating spatial resolution/uniformity and a section simulating a transverse section of the midbrain. The resolution/uniformity section is a cylinder 12.7 cm long, 20 cm diameter. Spatial resolution was evaluated subjectively with 2 sets of rods (“hot” and “cold”) of varying diameter (4.0, 5.0, 6.25, 7.81, 9.67 and 12.2 mm) and spacing (equal to 2X the rod diameter). If properly filled, the hot rods had a 8:1 target-to-background ratio. The clinical simulation section of the phantom was generated from patient MRI scans with the ventricles, gray and white matter separated into compartments and the gray and white matter sections fill to a ratio of 4:1. Uniformity and image quality was evaluated using the standardized uptake value (SUV) in a small volume of interest (V0I) as well as subjectively by 2 independent observers using a 4-point scale, and their results were then the averaged. RESULTS: The phantom was imaged on 13 PET scanners at 11 PBTC sites. The phantom’s complexity led to suboptimal filling by 5 of the 11 sites, particularly of the resolution hot rod section. Only 5 of 13 scanners reported an SUV in the uniformity section within 10% of unity although 12 of 13 were subjectively judged to have very good to excellent uniformity. For spatial resolution, 4 of 6 hot rods were discernible by all 13 scanners while 3 of 6 cold rods were discernible by only 5 scanners. For the clinical simulation section, only 4 of 13 scanners had a gray/white matter ratio between 3.0 and 5.0 (4.0 is truth); however, 11 of 13 scanners were subjectively judged to have very good or excellent image quality.

Conclusions: The SNMMI CTN brain validation phantom provided considerable interesting data across the PBTC PET scanners. However, challenges with filling and imaging this complex phantom led to variable data across sites.