Abstract
572
Objectives: Studies have shown that TOF gain in terms of SNR and lesion detectability is proportional to background size. However, most of the studies were performed on early generation of TOF scanners with effective timing resolution of 550 - 750 ps. TOF gain in terms of effectivity system sensitivity based on images reconstructed using OSEM also remains unclear. In addition, the dependence of TOF gain on lesion location in association with background size has not been thoroughly investigated. In this study, we performed phantom experiments to quantify TOF gain on a PET/CT scanner with approximately 400 ps timing resolution.
Methods: To quantify TOF gain on sensitivity improvement, we scanned a cylinder phantom (diameter 35 cm) with 9 spheres (diameters ranging from 8 mm to 37 mm) positioned in an inner and outer ring. A total of 2.1 mCi 18F-FDG was injected into the phantom and the sphere to background contrast was 8:1. Forty-minute listmode data was acquired and rebinned into 10 count levels (1-min - 10-min). Bootstrapping was used to generate 30 noise realizations from the 40-min data for each count level. Images were reconstructed using OSEM with 15 iterations and 10 subsets, with and without TOF. Contrast recovery coefficient (CRC) against background std was plotted for each sphere. Using 2-min TOF reconstruction at 3 iterations as a clinically relevant operating point, the two nearest data points on the non-TOF CRC curves (e.g. 9-min and 10-min) sitting above and below the operating point were averaged to calculate the sensitivity gain, i.e. (10-min + 9-min)/2/2-min = 4.75. This sensitivity gain was calculated for each sphere individually. To evaluate TOF gain for location and background size dependency, we developed a cross section (CS) phantom. The tapering phantom consists of 5 sections with varying diameters (20, 24, 30, 36 and 45 cm). Three fillable line inserts (diameter 10 mm) were fixed at 0, 4 and 8 cm radial offset positions, respectively, which ran through all the cross sections. A total of 8.9 mCi 18F-FDG was injected into the phantom and the line insert to background contrast was 4:1. All images were reconstructed using OSEM with 10 subsets and up to 50 iterations, with and without TOF information. CRC versus background coefficient of variation (CoV) curves were plotted for each section and each sphere.
Results: TOF reconstructions yield comparative image quality to those from non-TOF datasets acquired at an average of 4.1 times longer, which is equivalent to a factor of 4.1 increase in sensitivity. The CS phantom shows that the CRC improvement on the line inserts is both background and location dependent. Larger improvement is observed in bigger sections, and on the off-center inserts. It is also observed that TOF CRC curves reaches plateau at earlier iterations than the non-TOF curves, and yields more consistent CRC curves across different sections and inserts, which show that TOF information improves reconstruction convergence and the uniformity of the convergence in association with varying background size and insert location.
Conclusion: Our results show that TOF information with approximately 400 ps timing resolution can yield an average of 4.1 times sensitivity gain in terms of reconstructed image quality on the 35 cm cylinder phantom. TOF information can also increase convergence speed and uniformity of convergence especially for objects with large background, which might be particularly important for patients with large body dimensions. Research Support: Toshiba Medical