An investigation of statistical power of [15O]-H2O pet perfusion imaging: The influence of delay and time interval

Abstract

In the study of human brain functions, positron emission tomography (PET) makes signifi cant contributions in the form of “activation studies.” Activation studies with [ 15 O]-H 2 O exploit the phenomenon that activated brain regions have altered perfusion by blood. Statistical comparing of two conditions or groups on a voxel basis yields an activation map. Because blood delivery to the brain is spatially and temporally inhomogeneous, we hypothesized a strong dependencyof the activation map on the time periods that are compared. We aimed to systematically chart this infl uence in order to optimize sensitivity and time resolution. Methods: We dynamically scanned 11 volunteers with [ 15 O]-H 2 O while they performed a motor task and a control task. We localized statistical differences between the tasks with a statistical parametric mapping on all possible permutations of consecutive frame summations. Results: There was robust activation in the cerebellum and sensorimotor cortex and results were statistically most signifi cant for the period 20–50 s. Intervals as short as 10 s also yielded statistically signifi cant brain activations. The size of the activated clusters was strongly dependent on the interval that was compared and the optimal interval was different for the two locations. Conclusion: Cluster size was indeed strongly infl uenced by time interval and location in interaction, but cluster signifi cance appeared independent of location. Time periods of 10 s (20–30 or 30–40 s) were feasible, but maximal statistical power was obtained for 20–50 s durations. This knowledge allows maximization of sensitivity of [ 15 O]-H 2 O PET experiments through rational experimental and statistical design.