Sampling requirements for dynamic cardiac PET studies using image-derived input functions.

The utilization of image-derived input functions is becoming common in quantitative PET studies of the heart. Consequently, imaging protocols must be designed to sample both blood and tissue concentrations adequately. Most clinical imaging protocols consist of a series of short initial scans to measure the rapid change in blood and tissue tracer concentration levels, followed by scans of gradually increasing length. The number of initial short scans must be matched to the shape of the input function. In this paper, noise-free simulation studies were performed to evaluate the effect of temporal sampling on estimates of the parameters of a two-compartment kinetic model. In addition, the consequences of varying tracer infusion length and timing were studied. The kinetic model parameters' bias decreased when infusion times were lengthened or sampling rates increased. Our results indicated that tracer infusions of 30 sec were best suited for these studies. Two currently employed clinical imaging protocols were then optimized for use with this infusion scheme. Ten initial scans with durations of 10 sec, or twenty of 5 sec length produced unbiased estimates of kinetic model parameters that describe myocardial physiology. Noisy simulations with the equivalent of one million events confirmed these results.