📈 The haemodynamic response
Build a task and see the slow, delayed BOLD response the scanner will record.
The canonical HRF: one brief event
SPM's double-gamma curve. Real responses vary by region and person; the small "initial dip" some studies report is not part of this model.
task on / events
predicted BOLD (task convolved with the HRF)
what the scanner records, one point per TR
How to read this
- Top chart: the response to one brief event. It starts after a second or two, peaks about 5 s later, dips below baseline (the undershoot) and is back by about 25–30 s.
- Prediction: the expected BOLD signal is each moment of the task added up through the HRF (a convolution). Blocks merge into a plateau; events close together overlap and add up.
- TR: the scanner takes one whole-brain volume every TR seconds, so it only sees the orange points. Short TRs follow the curve's shape better.
- Noise and runs: real data are noisy. Averaging more runs (or more trials) shrinks the noise by the square root of their number: 4 runs halve it. The readout shows how well the recorded points match the prediction (correlation).
This prediction is what a GLM uses as a regressor: the model asks how strongly each voxel's time series follows it (lesson 15).
Key terms
- HRF
- Haemodynamic response function: the BOLD signal over time after a brief burst of neural activity.
- Convolution
- Sliding the HRF along the task timing and adding up the overlapping copies.
- Block design
- Task on for a block of seconds, then rest; big, robust signals.
- Event-related design
- Brief, separate trials with varied gaps; lets you study single events.
- TR
- Repetition time: seconds between two whole-brain volumes.
- Regressor
- A predicted time course used in the GLM to explain a voxel's signal.
Sources
- Friston K. J. et al. (1998). Event-related fMRI: characterizing differential responses. NeuroImage 7(1): 30–40 (the canonical HRF and its use in the GLM).
- Handwerker D. A., Ollinger J. M., D'Esposito M. (2004). Variation of BOLD hemodynamic responses across subjects and brain regions. NeuroImage 21(4): 1639–1651.
- Poldrack R. A., Mumford J. A., Nichols T. E. (2011). Handbook of Functional MRI Data Analysis, chapter 5. Cambridge University Press.
Used in lessons 5, 15 of fMRI from zero.