Guidance from F-18 FDG Vascular Lesion Image Simulations for the Design of Procedures for Vascular Lesion Quantification

Objectives: Quantifying vascular lesions with F-18-FDG at dimensions comparable to those of the PET image voxels presents serious difficulties arising from partial volume effects¹. To establish conditions under which the vessel lesion activity could be measured with acceptable certainty, we simulated time-activity profiles across large arteries with realistic lesions.

Methods: The model consisted of activity vs distance profiles across segments of an arterial vessel. Profiles were generated at several uptake time points and vessel dimensions. For the blood in the lumen, a population-based blood time-activity curve was used and a Patlak model provided the time-dependent activity of the lesion in the vessel wall. The simulations sought to reflect the activity that would be measured from an artery with a 2mm thick wall (such as the aorta at the level of the arch) as measured from a cross sectional view obtained from an image from a PET scanner having a 4mm FWHM point spread function [PSF] reconstructed on a 2mm x 2mm per voxel, 256x256 matrix. The true and the simulated voxel activities with no partial volume correction were determined for various uptake times following the injection of the tracer. Two related quantities were examined: contrast for the different wall and lumen activities and lesion contrast as a function of the uptake time.

Results: The simulations showed the large effect of the spill-in of blood activity on the measured wall activity, particularly at <60 min uptake time. As greater uptake time allowed higher clearance of blood, the FDG uptake of the wall lesions became more apparent. However, because of the small wall thickness of the vessel (2mm), even with no activity in the lumen, the recovery coefficient never exceeded 0.6 with a scanner with a PSF of 4mm FWHM. Since blood activity at the tissue/blood boundary can be accurately described as a step function, knowledge of the PSF allows the blood-to-wall spill-in to be estimated accurately. Because at the center of the lumen of a ~20mm vessel, 100% recovery was predicted by the simulations, this spill-in (blood into vessel wall) could be removed - assuming no significant wall motion.

Conclusion: Because the recovery coefficient for the spill-out at the wall depends only on the PSF and the thickness of the wall (assumed uniform radial uptake and no significance motion), the simulations showed that it is possible to accurately recover the actual lesion uptake from an activity profile without having to wait until the blood activity is minimal. This would avoid the need of acquisitions after long uptake periods which result in poor count statistics. Thus, the particular geometry and size of large vessels lend themselves to accurate lesion quantification provided that: 1) the PSF of the scanner be known; 2) an accurate estimation of the wall thickness is available (e.g. from contrast CT or MR images) and 3) that when motion is significant, that it be corrected through gating or by some other method. 1. Soret, M, Bacharach, SL and Buvat, I. Partial Volume Effect in PET tumor imaging. J Nucl Med 48 (6) 932-945, 2007 Research Support: This investigation was supported in part by the NIH Summer Internship Program and the Intramural Research Program of the NIH Clinical Center.