Abstract
459
Objectives: The new generation of digital whole-body PET scanners provide a combination of improved sensitivity due to increased axial length, as well as improved spatial and timing resolution. In particular, the Siemens Biograph Vision provides a desirable combination of high sensitivity (15.1 cps/kBq), spatial resolution (3.6 mm), and timing resolution (210ps) that is expected to lead to improved clinical performance for small lesion detection with short imaging times. However, translation of these improved physical characteristics into clinical image quality will also be determined by the image reconstruction algorithm and in particular any filter that may be used for post-reconstruction smoothing. This work performs a small lesion detectability study on the previous and latest generation of Siemens PET/CT scanners, the Biograph mCT and Vision, to study the impact of improved physical performance and the choice of image reconstruction method on the clinical performance as a function of imaging time.
Methods: Measurements were performed separately on the mCT and Vision scanners with the SNMMI CTN phantom filled with 0.16 uCi/cc of 18F-FDG. List-mode data were acquired for 60 minutes in order to obtain multiple replicate data sets of shorter scan durations. Lesion data acquired separately using spheres in air of varying size and uptake were combined with the CTN phantom data to generate list data with embedded lesions. Image reconstruction was performed for a range of imaging times. Standard image reconstruction was performed using the ordinary Poisson ordered subsets expectation maximization (OP-OSEM) algorithm with PSF+TOF modeling and 5 mm post-reconstruction Gaussian filter. The current choice of post-reconstruction filtering for the Vision was made to maintain continuity in qualitative image quality for patients previously scanned on mCT. In order to better leverage the improved performance of Vision we also evaluated images without application of a post-reconstruction filter. We used a generalized scan statistics methodology to estimate the ALROC as a function of scan time and choice of filtering to determine the impact of the improved physical performance of Vision on lesion detection and localization with reduced patient imaging times.
Results: Current measurements using the NEMA Image Quality and CTN phantoms indicate higher contrast recovery with the Vision scanner and potential to improve lesion detectability. Our previous simulation results indicate almost a factor of two gain in ALROC values (or a commensurate reduction in imaging time) with improved spatial and timing resolutions and we expect these to translate to our measurements. With the higher sensitivity of Vision we also expect further reduction in background noise leading to additional gains in the ALROC values as the imaging time is reduced.
Conclusions: Our experimental measurements using a numerical lesion localization and detectability metric quantify the gain in image quality that is achieved with the new Vision scanner. These results indicate improved clinical performance for detection of small lesions with short imaging times.Research Support:This work was supported in part by the National Institutes of Health under Grants R01-CA113941 and R01-CA196528 & by a research agreement with Siemens.