Simultaneous monitoring of PET image resolution, noise, uniformity and quantitative accuracy using uniform cylinder phantom measurements in the multi-center setting

We have proposed and implemented simultaneous monitoring of image resolution and noise, as well as image uniformity and quantitative accuracy, in a multi-center study of vascular dementia, only utilizing cylinder phantoms. Three centers were enrolled within the ARIC (Atherosclerosis Risk in Communities)-PET study involving amyloid imaging using F-18 Florbetapir. Weekly uniform phantom scans involving ∼1mCi of F-18 were conducted for the study duration in each of the three centers. The acquisition protocol consisted of 4×5min frames, consistent with the subject scans (4×5min, 50-70min post-injection). Image resolution was extracted by analyzing the blurring extent of the phantom edges. Specifically, the effective point-spread function (PSF) was obtained by taking the derivative of edges, followed by fitting with a Gaussian function. Noise was extracted by analysis of two back-to-back dynamic images, subtracted from one another, to remove the impact of any spatial non-uniformity, followed by computation of the standard deviation, which was then corrected by sqrt(2) given the difference operation. The phantom was also used to assess uniformity, using ratio of mean uptake within an inner circle to uptake within an outer annulus. Having knowledge of the exact dose, times of injection and scan, as well as phantom volume, enabled measurements of the true concentration, which in turn allowed assessment of quantitative accuracy. Our proposed framework enables feasible and routine multi-center monitoring of image quality, which also allowed us to arrive at a number of interesting observations and conclusions.