Investigation of four phantoms for PET normalization

Normalization in Positron Emission Tomography (PET) is essential to correct for sensitivity variation across detector elements and produce uniform images. Currently, the standard normalization protocol for General Electric (GE) systems requires a 20 minute scan of a 20 cm diameter cylindrical phantom filled with 2 mCi of 18F Fluorodeoxyglucose (FDG). This process calculates the single detector efficiency factors (e) from the component based normalization factors (NF). We have investigated the feasibility of using smaller phantoms for normalization, including sensitivity to positioning and radioactivity concentration. Four new test phantoms were used for normalization on a GE Discovery 690 PET/CT scanner. The test phantoms included 4.5 cm, 7.5 cm and IOcm diameter cylinders, and a 16-20 cm diameter cylindrical shell. Normalization scans were also taken using the 7.5 cm and 10 cm cylinder phantoms over a range of FDG activities and at transaxially off-center positions. Image quality was assessed by reconstructing a 4 hr high count, uniform activity image quality (IQ) phantom scan using each normalization data set. All test phantoms produced similarly uniform reconstructions of the IQ phantom scan. Upon examining subtractions from the standard, high frequency ring artifacts with ±1 % peak intensity were present in images reconstructed using the 4.5cm, 7.5cm, and 10cm normalization phantoms. FFT analysis quantified the spatial period of ring artifacts as T = 2.15 ± 0.1 cm, one half of the detector block width of the GE 690 system. Simulation confirmed that ring artifacts result from periodic detector sensitivity decrease, which points to incident photo obliquity being the cause of said artifact. ROIs across anterior/posterior direction of the central slice were used to quantify low frequency uniformity. Low frequency non-uniformities were found to be produced by off-center normalization phantom positioning, as expected. In addition, an 8 minute Hoffman brain phantom scan was reconstructed using each a centered normalization data set from each phantom. Grey matter was isolated and statistically assessed using mean and max intensity difference from the standard. When subtracted from the standard reconstruction, all four test phantom images displayed ~0.5% standard deviation of pixel intensity variation, which did not create a visibly noticeable difference in the reconstructed images.