Abstract

Most fMRI studies assume a shape for the underlying hemodynamic response function(HRF) when they construct models of predicted fMRI signals. Aguirre, et.al.(1998) showed that the shape of the HRF varies greatly among subjects, but is repeatable within subjects and across sessions. This highlighted the potential need to empirically derive an HRF for each subject. Miezin, et.al.(2000) demonstrated that some parameters describing the hemodynamic response in subjects' visual cortices cannot predict the same parameters in motor cortices. If HRFs in one part of the brain cannot be used to model an HRF in another part of the brain, then an empirically derived HRF for each subject might be no better or worse than a generic HRF used for all subjects. Subjects participated in a simple visually guided motor and oculomotor task that activates visual cortex, motor cortex, frontal eye fields, and supplementary eye fields. An HRF was empirically derived from each region for each subject. Each HRF was fitted to the sum of two gamma functions to find parameters such as time-to-peak and amplitude, which define the HRF. These parameters were compared among regions and among subjects. Additional analyses examined the inter-region and inter-subject variability to determine if an empirically derived HRF yields more accurate results than a generic HRF. Results show that although there is variability among the regions within each subject, there is more variability across subjects. This indicates that in many instances the use of an empirically derived HRF can increase the statistical power of fMRI experiments.

From: Handwerker, D. A., J. M. Ollinger, et al. (2002). Effects of regional and subject variability of the hemodynamic response function on modeling fMRI signals. Society for Neuroscience, Orlando.