The purpose of this study is to establish and validate a methodology for estimating the standard deviation of voxels with large activity concentrations within a PET image using replicate imaging that is immediately available for use in the clinic. To do this, ensembles of voxels in the averaged replicate images were compared to the corresponding ensembles in images derived from summed sinograms. In addition, the replicate imaging noise estimate was compared to a noise estimate based on an ensemble of voxels within a region. To make this comparison two phantoms were used. The first phantom was a seven-chamber phantom constructed of 1 liter plastic bottles. Each chamber of this phantom was filled with a different activity concentration relative to the lowest activity concentration with ratios of 1:1, 1:1, 2:1, 2:1, 4:1, 8:1 and 16:1. The second phantom was a GE Well-Counter phantom. These phantoms were imaged and reconstructed on a GE DSTE PET/CT scanner with 2D and 3D reprojection filtered backprojection (FBP), and with 2D- and 3D-ordered subset expectation maximization (OSEM). A series of tests were applied to the resulting images that showed that the region and replicate imaging methods for estimating standard deviation were equivalent for backprojection reconstructions. Furthermore, the noise properties of the FBP algorithms allowed scaling the replicate estimates of the standard deviation by a factor of 1/square root N, where N is the number of replicate images, to obtain the standard deviation of the full data image. This was not the case for OSEM image reconstruction. Due to nonlinearity of the OSEM algorithm, the noise is shown to be both position and activity concentration dependent in such a way that no simple scaling factor can be used to extrapolate noise as a function of counts. The use of the Well-Counter phantom contributed to the development of a heuristic extrapolation of the noise as a function of radius in FBP. In addition, the signal-to-noise ratio for high uptake objects was confirmed to be higher with backprojection image reconstruction methods. These techniques were applied to several patient data sets acquired in either 2D or 3D mode, with (18)F (FLT and FDG). Images of the standard deviation and signal-to-noise ratios were constructed and the standard deviations of the tumors' uptake were determined. Finally, a radial noise extrapolation relationship deduced in this paper was applied to patient data.
Purpose: Our goal is to investigate the effect on PET accuracy of the various physical and instrumental phenomena using realistic Monte Carlo (MC) models of patient activity distributions and PET scanners. Methods and materials: Data from a lung PET/CT scan is used to create numerical voxelized phantom for realistic MC simulations using the GATE MC code. We compared the activity reconstructed from the simulations to the input (“true”) values. We investigated the effects of random coincidences, photon scatter and positional shift between transmission and emission scans on the degradation of 3D PET quantification accuracy. Results: When exact attenuation, scatter and random corrections are applied, the activity from the simulated scans deviated from the input by more than 50% for 25% of the voxels. Inaccuracy of the random and scatter corrections of 20% will add additional quantification errors of about 20% and 40% respectively. The attenuation correction (AC) adds random and scatter events to the high tissue density regions. AC based on misaligned transmission data causes pronounced changes in the activity distribution in regions that have a high density gradient such as at junctions between soft tissue, bone and lung. AC misalignment causes errors of the activity levels, which may exceed a factor of five around the ribs for 1.5 cm shift between the PET data and the AC image. Conclusions: MC methods are a useful tool in investigating PET scanner response to realistic distributions of the activity and the attenuation properties and in establishing the uncertainty in PET quantification in patient studies. The potential quantification errors due to possible misregistration and inaccurate scatter and random corrections have been determined as indicated above for the lung. Supported in part by NCI grant CA‐059017 and performed in part on the Physon cluster, funded by Bulgarian NSF contract DO‐1‐872.
Purpose: The needs of radiation therapy treatment planning impose higher demands on PET/CT imaging accuracy. The recently developed GATE (Geant4 Application for Tomographic Emission) Monte Carlo package, provides the possibility to model accurately the factors contributing to decreased PET resolution and image degradation. The purpose of this study is to test GATE's ability to predict time curves and image quality (IQ) for the GE Discovery LS/Advance PET scanner. Method and Materials: Our 3D PET simulation model of the GE Discovery LS scanner and phantoms follows both the vendor's and NEMA's specifications and was previously validated for the PET scatter fraction and sensitivity tests. Simulations with this model were performed for the count rate and IQ NEMA‐2001 tests as a function of activity concentration. The Software for Tomographic Image Reconstruction (STIR) package was used to reconstruct the simulated data which was then compared to experiment. Results: Our simulations correctly predict the shape and magnitude of the true, scatter, random, and NEC rates. The simulated peak true and NEC rates are both within 3 kBq/cc of the measured data. Scatter and random rejection in the Monte Carlo data dramatically improved the agreement between measured and simulated contrast ratios. The cold sphere contrast ratio is within 9% of the measured data when the scatter and random coincidences were rejected. Larger discrepancies for the hot sphere contrast are currently observed and investigated. Conclusion: Monte Carlo simulation of PET images and the corresponding NEC rates can aid in improving PET image quality. The ability to model the features of PET scanners accurately makes GATE a potentially useful tool in improving PET's performance, which is necessary for its effective use in radiation treatment planning This work is sponsored by Program Project Grant CA059017‐12 from the National Cancer Institute and by an internal departmental grant.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.