Abstract
3291
Introduction: Selective internal radiation therapy (SIRT) is increasingly used to treat liver cancer and shows potential to improve quality of life. However, the success of the therapy relies on the tumor response, which is reported by a growing number of studies to be linked to the absorbed dose. Imaging the 90Y activity distribution with positron emission tomography (PET) is currently the best option for computing the liver dosimetry since, despite the count-starved images due to very low positron emission (~32 ppm), it outperforms bremsstrahlung imaging routinely done with SPECT. However, since quantification accuracy and noise properties of the images are largely affected by the iterative image reconstruction protocol, variations in the estimated absorbed dose can be expected when PET reconstructed images are used for dose estimation. The aim of the present study was to quantify these dosimetric variations caused by different 90Y image reconstruction protocols using the NEMA IQ phantom, and to provide a direct comparison of these effects when conventional and total-body PET scanners are used.
Methods: A NEMA NU 2 image quality (IQ) phantom was used to evaluate the absorbed dose in different volume sizes. The phantom consisted of a warm background compartment and 6 spheres (diameters 10-37 mm) filled with 2.51 GBq of 90Y solution for a sphere-to-background ratio of 7.78:1. The phantom was positioned with all spheres at the center of the axial field-of-view of each scanner and imaged for 30 min using a single bed position on both the Biograph mCT (Siemens Healthineers) and the uEXPLORER PET/CT (United Imaging Healthcare).Different image reconstruction protocols were investigated for the uEXPLORER and the mCT to quantify the effects from voxel size and using the point spread function modeling. The number of iterations and subsets was kept constant to 4 iterations and 20 subsets for the uEXPLORER and 3 iterations and 21 subsets for the mCT.The absorbed dose was calculated by Monte Carlo simulations using Geant4 Application for Tomographic Emission (GATE) version 9.0. Both the CT and PET images were used as inputs to the simulations to provide the phantom geometry, including material composition and density, and PET-derived activity distribution defined by the heterogeneous 90Y β- source. For each image reconstruction, the NEMA IQ phantom spheres were simulated individually for easy scaling of the absorbed dose, given by the cumulative activity from the simulation performed with a fixed number of primary events of 107. A reference absorbed dose distribution was also generated with Monte Carlo simulations and a numerical phantom matching the physical phantom spheres' dimensions.
Results: The dose varied largely depending on the sphere size, the reconstruction parameters, and the scanner. The relative disagreement, depending on the sphere size, was between 1.7 Gy to 5.7 Gy and 0.7 Gy to 2.4 Gy for the uEXPLORER and mCT, respectively. When compared to the reference absorbed dose, both scanners underestimated the dose for the 10 mm sphere (Figure 1A), with an underestimation of up to 7% with the uEXPLORER and 12% with the mCT. For the 37 mm sphere (Figure 1B), the two scanners overestimated the absorbed dose, but by a smaller difference from the ground truth (2% or less, depending on the voxel size). Image voxel size seemed to have little impact on the estimated dosimetry, while the use of PSF modeling increased the absorbed dose in all reconstructions, except for the 10 mm sphere, leading to an overall increased absorbed dose estimate.
Conclusions: Different reconstruction protocols resulted in variations on the absorbed dose with a larger impact on small volumes. Reconstruction parameters should be carefully investigated and chosen in order to produce images with activities that closely match the injected activity and provide a reliable absorbed dose estimation.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.