The range of experimental design strategies for PET is limited by the physical constraints associated with delivering and measuring radioisotopes. At a minimum, 150 studies involve an aggregation of activity occurring over 30 seconds or more, and for metabolism and neurotransmitter studies the aggregation covers 20 to 60 minutes. This temporal resolution precludes event-related types of designs and makes it difficult to dissociate the different processes involved in a task. The total number of radiotracer injections (and hence scans) is limited by radiation exposure and the need to allow previously administered radiotracers to decay substantially before starting the next scan (this takes hours for nC and 18F and about 8 to 10 minutes for 150). With 150, one is typically limited to about 12 scans in a 2-hour scanning session, and with 18F and nC, one is typically limited to 2 to 4 scans (usually scheduled on different days because the subjects would need to spend hours waiting for the isotope to decay between scans). However, because the data in each scan is an aggregate of activity over time, a single contrast between two PET scans can be informative, whereas a contrast between two individual BOLD images (one phase each) has little value.
The most common PET analysis involves a simple subtraction paradigm. As with block designs in fMRI, these analyses are efficient and straightforward, but often depend on the problematic assumption of pure insertion. Parametric designs, where a variable is quantitatively manipulated across different conditions, and factorial designs are also frequently implemented. All of the preceding designs can be analyzed within the framework of the general linear model (Friston et al., 1995). Studies examining the covariance of activity across regions and application of techniques to assess functional connectivity are also possible, although the statistical power to apply such techniques is often restricted by the limited number of scans (Friston, Frith, Liddle, & Frackowiak, 1993; Zald, Don-ndelinger, & Pardo, 1998).
Was this article helpful?