A new PET image quality phantom, acquisition and analysis paradigm

Timothy Turkington; John Sunderland; Yanic Bercier; Jeffrey Yap; R Glenn Wells; Christian Schütze; Mengxi Zhang; Stefan Siegel

Introduction: The NEMA/IEC PET image quality method has been a primary standard for over 2 decades for system specifications, acceptance testing, quality control, scanner qualification and other tasks. Continuous advances in PET hardware have driven the need for a modified, more challenging image quality phantom, and an unbiased acquisition and analysis methodology that accounts for length of the axial Field of View (aFOV). The proposed phantom, acquisition protocol, and analysis methodology have been adapted from the original NEMA/IEC paradigm, but with some notable changes. The phantom and analysis methods were assessed in a test/retest protocol on two different PET systems, for reproducibility of contrast recovery coefficient (CRC) and background variability (BV), using a previously described modified NEMA phantom (Turkington et al. SNMMI 2024). The phantom includes 6, 8, 10, 13, 17, and 22mm diameter spheres in a central ring of 57mm radius, three 6 and three 8mm spheres at a 114mm radius. Two of the three outer 6mm spheres, and two of the three outer 8mm spheres are offset by + and 1mm. All spheres have an activity concentration 4X background. Methods: The new methodology utilizes a single bed-position 40-min. acquisition with the central scanner slice aligned with sphere centers. A single 40-minute high statistics image was generated. The list mode data was subsequently parsed into multiple short duration realizations with lengths designed to be consistent with a typical clinical acquisition that will allow for fair comparison between scanners with different aFOVs and bed overlap. To generate a level playing field, the acquisition replay paradigm was designed to simulate image quality associated with a whole body scan with duration TWB=20 minutes that covers a standard 100 cm of scan length (R) for a scanner with aFOV = A. The duration of the aFOV specific list-mode replay phantom noise realization time Tphant is Tphant = Save/Sspheres * TWB* min(A/R,1), where the term min(A/R, 1) accounts for long aFOV scanners greater than 100cm. The 40 min. acquisition is divided into as many Tphant duration frames as possible. The ratio Save/Sspheres accounts for axial sensitivity profile of the scanner, where Save and Sspheres are the average slice sensitivity and the sensitivity at the central slice. Volumes of interest (VOIs) for the spheres and background regions were specified in a manner consistent with the NEMA NU 2-2024 procedures, with the background VOI placement adapted to avoid the additional features of the modified phantom. The results from each frame are averaged to produce the final metrics for the system. Results: Two systems of 16 cm (SYS16) and 106 cm (SYS106) in aFOV dimension were each tested on 2 different days to verify reproducibility of CRC and BV. The reconstructed images from the 40 minute data were super-sampled by a factor of 10 in each direction. The centroids of the 12 sphere-centers were identified using an automated software analysis program written in MATLAB. The methodology resulted in 24 Tphant = 1min38sec realizations for SYS16 and 4 Tphant = 10min0sec noise realizations for SYS106. See Figure 1 for results. Importantly, CRC and BV measurements for the new 6mm and 8mm spheres were highly reproducible using the prototype phantom, phantom fill and acquisition protocol. Conclusions: A new PET image quality method has been designed and tested to support current and future systems for all aFOV tomographs and overlap settings. The new prototype image quality phantom with spheres ranging in diameter from 6 to 22mm was used for this study. Preliminary results indicate that the new method is reproducible in multiple tests, for both short and long aFOVs. Work is ongoing to expand the number of systems included in the analysis.

A new PET image quality phantom, acquisition and analysis paradigm

View Online

Abstract

Details

Metrics