Effects of scatter on model parameter estimates in 3D PET studies of the human brain

Abstract

Phantom measurements and simulated data were used to characterize the effects of scatter on 3D PET projection data, reconstructed images and model parameter estimates. Scatter distributions were estimated from studies of the 3D Hoffman brain phantom by the 2D/3D difference method. The total scatter fraction in the projection data was 40%, but reduces to 27% when only those counts within the boundary of the brain are considered. After reconstruction, the whole brain scatter fraction is 20%, averaging 10% in cortical gray matter, 21% in basal ganglia and 40% in white matter. The scatter contribution varies by almost a factor of two from the edge to the center of the brain due to the shape of the scatter distribution and the effects of attenuation correction. The effect or scatter on estimates of cerebral metabolic rate for glucose (CMRGl) and cerebral blood flow (CBF) is evaluated by simulating typical gray matter time activity curves (TAC's) and adding a scatter component based on whole-brain activity. Both CMRGl and CBF change in a linear fashion with scatter fraction. Errors of between 10 and 30% will typically result if 3D studies are not corrected for scatter. We also present results from a simple and fast scatter correction which fits a gaussian function to the scattered events outside the brain. This reduced the scatter fraction to <2% in a range of phantom studies with different activity distributions. Using this correction, quantitative errors in 3D PET studies of CMRGl and CBF can be reduced to well below 10%.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>