| |
Abstract:
Event-related fMRI designs promise to be a powerful method
for studying the neural mechanisms of cognition. In these designs,
we wished to investigate the theoretical dependence of sensitivity
(to underlying neural effects) on the temporal design parameters of
inter-trial interval (ITI) and the probability of a trial actually
occurring per ITI (p). We also wished to compare the sensitivity of
blocked designs to that of event-related designs. Methods: The fMRI
signal was modeled as the output of a linear, time-invariant
Gaussian hemodynamic transform of neural activity plus (frequency-1
+ uncorrelated) noise. This model allowed the derivation of
signal:noise expressions that depended on ITI and p. Results: For
evenly-spaced (i.e., p=1) event-related designs, there is a global
maximum in signal:noise at ITI=2ps seconds, while for 0
|
