Spatially dependent deadtime losses in high count rate cardiac PET

Cardiac PET scans result in nonhomogeneous distributions of activity within the body, which might lead to great variations in singles rates around the detector ring. Conventional deadtime correction algorithms assume that the singles rates are uniform. This paper investigates singles nonuniformities during several typical cardiac scanning protocols (bolus injections of 15O-water and 82Rb, slow infusion of 18F-FDG and static imaging with FDG) and estimates how such nonuniformities might affect quantitative data. Nonuniformity was observed in all studies and was described by an asymmetry index which increased to 58% during bolus water injection, the most inhomogeneous study. These results are valid for any scanner with a ring diameter of approximately 78 cm and are independent of the amount of activity injected. Deadtime losses depend on the amount of activity and on the scanner type. Nonhomogeneities in singles can be shown to produce spatially dependent deadtime correction factors; for our scanner, these were seen to differ by up to 16% from the mean deadtime correction during bolus water injection. To demonstrate the distortions generated by average deadtime correction, the activity distribution during a clinical cardiac study was simulated using a phantom. A simple local deadtime correction and its implementation on our system are described, and the resulting improvements in both absolute and relative quantitation of the phantom study are shown.