HRF Papers

Useful Papers - Your Friend, The Hemodynamic Response Function


Aguirre et al. (1998), "The variability of human BOLD hemodynamic responses," NeuroImage 8, 360-369 PDF

Summary: In order to evaluate how much variability exists in the shape of HRFs collected from different sessions, days, and subjects, Aguirre et. al tested various sets of 40 subjects tested on various days and in various sessions. They found a good deal of variance accounted for by differences in subjects, and significant differences for many subjects between different scanning days. Within the same day and subject, thought, the HRF seemed relatively stable.

Bottom line: Shows that subject-to-subject and day-to-day variance in HRF can be high, but within a day across runs, the HRF is relatively stable.


Miezin et al. (2000), "Characterizing the hemodynamic response: Effects of presentation rate, sampling procedure, and the possibility of ordering brain activity based on relative timing (Ed. - whew!)," NeuroImage 11, 735-759 PDF

Summary: A long paper describing a number of different studies/analyses looking at linearity of the HRF (as stimulus presentation rate and sampling procedure varies) and the variance between regions, subjects, and different aspects of the HRF (amplitude, time-to-onset, etc.).

Bottom line: Excellent look at the factors influencing the HRF and how stable all its aspects are. Demonstrates HRF remains stable and linear within a subject and certain timing parameters, but outside those, less so.

Della-Maggiore et al. (2002), "An empirical comparison of SPM preprocessing parameters to the analysis of fMRI data," NeuroImage 17, 19-28 PDF

Summary: Another find-the-best-analysis-parameters study, this time using Monte Carlo simulations on simulated data to study power and false-positive rate as various SPM parameters changed. Of particular interest here is the contrast between using the canonical HRF alone vs. the HRF with temporal derivative. The HRF with temporal derivative is found to significantly reduce power in many circumstances.

Bottom line: The canonical HRF by itself does better than it does with the temporal derivative added in much of the time.

Neumann et al. (2003), "Within-subject variability of BOLD response dynamics," NeuroImage 19, 784-796 PDF

Summary: Neumann et. al investigate how variable the HRF is within a subject, over several scanning sessions spanning days or weeks. They look at stability of a number of parameters - time to onset, time to peak, time to baseline, etc. and use a couple different methods to extract HRFs. Time-to-peak is the most stable parameter they found, and voxels activated in all sessions had much lower variability than those only activated in some.

Bottom line: At least a subset of activated voxels within a subject has quite a stable hemodynamic response, with time-to-peak being the most stable aspect.


Schicke et al. (2006), "Tight covariation of BOLD signal changes and slow ERPs in the parietal cortex in a parametric spatial imagery task with haptic acquisition," European Journal of Neuroscience 23, 1910-1918 DOI

Summary: Participants learned an object map by touch and then imagined discs flying from object to object during ERP and FMRI acquisition. ERP amplitude parametrically varied with disc-flying distance, as did BOLD in left parietal cortex; source modeling suggested the ERP originated from that left parietal source.

Bottom line: Slow ERPs, thought to arise from postsynaptic activity, correlated with BOLD in parietal cortex, suggesting the BOLD response may reflect increased input to an area rather than spiking activity directly.

Logothetis & Wandell (2004), "Interpreting the BOLD signal," Annual Review of Physiology 66, 735-69 DOI

Summary: Comprehensive and excellent review of the complete chain of events in the head leading to the BOLD signal, including a bit of physics and a lot of neurophysiology. Several studies containing both cellular and BOLD recordings are reviewed, as are studies of neural vasculature and other regional differences influencing the HRF, and some suggestions are made for interpreting the BOLD signal.

Bottom line: BOLD correlates somewhat with summed local spiking, but somewhat better (and in places a great deal better) with lower-frequency electrical activity (often postsynaptic) summed over many neurons. In other words, BOLD results from a combination of pre- and post-synaptic activity - that is, outputs from an area and inputs into it - and probably correlates better with the input side. Comparing absolute BOLD amplitude between regions isn't justified, theoretically.

Liu & Newsome (2006), "Local field potential in cortical area MT: Stimulus tuning and behavioral correlations," Journal of Neuroscience 26, 7779-7790 DOI

Summary:LFP and multi-unit activity were extracted simultaneously from speed- and direction-sensitive areas of MT to try and get a handle on the spatial sensitivity of the LFP signal; presumably if its spatial resolution is very low, it wouldn't exhibit the same kind of direction- or speed-sensitivity found at the single- or multi-unit level. LFP activity did exhibit both speed and direction tuning, but only at higher frequencies (gamma band), suggesting that spatially localized information is mostly carried in the upper bands. Speed tuning was present only at higher frequencies than direction tuning, perhaps reflecting the finer-grained spatial organization of speed-tuned neurons.

Bottom line: The LFP (which, as noted above, contributes to BOLD and probably correlates better with itthan local spiking) has good spatial resolution in higher frequencies, at least to the level of cortical columns, suggesting that BOLD driven by LFP signals is not being contaminated by long-distance signals.

Robinson et al. (2006), "BOLD responses to stimuli: Dependence on frequency, stimulus form, amplitude, and repetition rate," NeuroImage 31, 585-599 DOI

Summary:Pretty mathematical paper using current models of the BOLD-neuronal firing relationship and parameters derived from ERP literature to derive the theoretical BOLD response for various shapes of neuronal responses (damped sinusoids, like ERPs). Also derives optimal block-stimulation frequencies.

Bottom line: For short stimuli, ,maximum BOLD peak is roughly proportional to the time integral (signed area under the curve) of neuronal activity. When neuronal response shapes differ (i.e., oscillate more or less), you might have higher firing-rate peaks for one stimulus but lower BOLD if area under the curve is different (due to oscillating firing rates). As well, to maximize BOLD response, a 7-sec-on, 7-sec-off block frequency is suggested based on the derived BOLD response.