Since the initial introduction of techniques for in vivo measurement of blood flow in humans in the 1980's, positron emission tomography (PET) techniques have steadily grown in popularity for answering neuroscientific questions about brain function in humans. Analysis and interpretation of blood flow studies requires a detailed understanding of the relationships between brain activity, blood flow, tracer distribution, the scanning environment and physical characteristics of the scanner. Simplifying approximations are often made in the interest of practicality or subject comfort. The most common statisticalm methods currently in use are based on standard analysis of variance techniques with adjustments to compensate for the presence of multiple correlated spatial comparisons. These techniques have been progressively refined and now utilize multi-way analyses of variance to allow identification of main effects of task, learning effects, and interactions among task components. Functional imaging studies are typically based on a small number of subjects, and as a consequence, non-sphericity associated with repeated measures is often routinely ignored. The small number of subjects is even more problematic when a patient group is compared to a control group. In this setting, subject identity is often treated as a fixed effect, raising doubts about the validity of inferences regarding the populations from which the subjects were drawn. Even when differences between patients and controls can be identified, these sometimes subsequently prove to be related to systematic anatomic differences between the groups rather than to true functional differences. Further refinement of statistical methods for analysis of PET blood flow studies will be important in maximizing the quality of the conclusions drawn using this technique.

Back to Workshop Schedule