Investigation of acquisition parameters for PVI normalization

The authors studied the effects of count density, count rate, and source geometry on a model based normalization using a GE Advance PET system. To study precision vs. count density the authors generated crystal efficiencies, ε, from 20 time equivalent PVI scans acquired with a 50 μCi rotating ⁶⁸Ge line source, and calculated the standard deviation, σ, of the individual ε over the 20 realizations. The σ followed the Poisson estimate of 1/√N where N is the total number of counts involving a given crystal. For example, an average value of 2.7 counts per line of response in the central 22 cm field of view resulted in 5,000 counts per crystal and a σ of 1.5%. Using a 20 cm cylinder filled with F-18, the authors studied count rate effects by measuring the difference between ε generated from acquisitions ranging in block deadtime (BDT) from 0.6% to 11.4% and those from a low rate, high count reference scan. Rate induced errors increased linearly from 0.5% to 4.8% over the 1.2% to 11.4% BDT range. The authors studied cylinder, annulus (20 cm OD and 14 cm ID), and rotating rod ⁶⁸Ge source geometries. Eight sets of ε were generated for each geometry, and compared cylinder to annulus and cylinder to rotating rod using the Student´s t statistic. Elliptical phantom data were normalized by each source geometry and reconstructed using 3DRP