Abstract
1410
Objectives For quantitative PET measurements to be comparable between different scanner systems, the reconstruction parameters have to be carefully optimized to produce images with similar spatial resolution. Various specialized phantoms reflecting different clinical settings have been developed for protocol optimization but these phantoms are not well suited for widespread deployment across the imaging community. This project aims to develop a simple phantom procedure that can be used to characterize spatial resolution and to compare these measurements with recovery coefficients obtained using the NEMA image quality phantom.
Methods The proposed method uses information from the edge of a cylinder phantom to measure the spatial resolution corresponding to a particular clinical protocol. A 20 cm diameter uniform 18F cylinder phantom was centrally positioned at a small angle (~5 degrees) with respect to the z-axis of a Biograph mCT. The oblique angle ensures the phantom edge intersects the image matrix differently in different slices, allowing very fine sampling of the edge spread function without having to alter the clinical reconstruction protocol (a practical advantage for multi-center deployment). Spatial resolution was measured as the full width at half maximum (FWHM) by fitting a model to the edge spread functions in both radial and axial directions. Support for the method was obtained by acquiring multiple phantom replicates to assess repeatability; controlled modulation of image reconstruction parameters; comparison with NEMA spatial resolution measurements (point source in air); and comparison with the NEMA image quality phantom at 2, 4, 6, 8 and 10-to-1 sphere-to-background ratios.
Results The table shows the mean FWHM ± 1 SD, confirming high repeatability of the cylinder method (1-3 % coefficient of variation). Corresponding NEMA point source measurements are also shown for comparison. The FWHMs measured from the cylinder were used to calculate recovery coefficients corresponding to the spheres in the NEMA image quality phantom by filtering an idealized digital reference object. The recovery coefficients derived using the cylinder FWHMs were in very close agreement with the recovery coefficients derived from the physical phantom (linear regression slope = 1.016, R2 = 0.96) over a range of sphere-to-background ratios and different reconstruction protocols.
Conclusions The edge of a uniform cylinder phantom can be used to estimate the spatial resolution that is achieved with clinical reconstruction protocols, as opposed to the limits of performance that can be obtained under optimized conditions. FWHMs derived using the cylinder method were highly predictive of recovery coefficients in an extended phantom with fillable spherical inserts. The cylinder method is well-suited for multi-center deployment as the phantom is low-cost, widely available, simple to prepare and easy to set up. In addition the proposed resolution measurement can be performed in conjunction with tests of calibration accuracy and will hopefully contribute to greater standardization of PET data collection across sites.