Problems in estimating hemodynamic response parameters from fMRI data.

The temporal (time-analysis) component of the statistical parametric mapping of the method of Friston et al. ([1994]: Human Brain Mapping 1:153-171) is evaluated using stimulated time-course data sets and images acquired during a motor-task activation study. The results indicate that their method for estimating hemodynamic delay parameters is unreliable when the hemodynamic delay time is greater than 1-2 image frames. Problems also arise when system noise is present in addition to physiologic noise. These conditions apply for most task-activation studies with image acquisition rates (TRs) shorter than about 3 sec, and for images acquired at field strengths below 3-4 tesla. Modifications are suggested that may make their method usable under these conditions. Copyright (c) 1997 Wiley-Liss, Inc.