Virtual laboratory / Brain regions and the whole brain
fMRI experiment design
The timing of an fMRI experiment and its effect on the power to detect brain activity.
Guided experiment: Mapping brain activity with fMRI.
fMRI experiment design
Functional MRI measures changes in blood oxygenation that follow neural activity. The haemodynamic response is slow, peaking about five seconds after a brief event, so the timing of an experiment determines how well responses can be estimated. Analyses usually model the signal as the stimulus timing convolved with a haemodynamic response function and fit a general linear model (Friston et al., 1998). Randomising the intervals between events greatly improves the efficiency of event-related designs with short intervals between events (Dale, 1999).
Design the experiment
Questions to investigate
- Compare a block design with an event-related design of the same length. Which detects the effect more easily?
- In an event-related design with a fixed interval, shorten the interval to two seconds. Then add jitter. What happens to efficiency?
- Halve the run length. How does the statistical power across participants change?
- Compare two conditions directly rather than each against rest. Why is this contrast harder to detect?
- Remove the high-pass filter. How does slow drift affect the result?
In a real laboratory
Scanner time is expensive, so designs are usually tested by simulation before data are collected. Participants are screened for MRI safety, give written consent and practise the task beforehand. Head motion is minimised and measured. Whole-brain analyses must correct for the many tests performed, which requires stronger effects or more participants than a single-region analysis. Research with people requires approval by an ethics committee (in India, an Institutional Ethics Committee following the ICMR National Ethical Guidelines of 2017; elsewhere, the equivalent body).
How the simulation works
The simulation models one brain region. The stimulus timing is convolved with a canonical double-gamma haemodynamic response function, scaled so that the peak response to a 1-s event equals the chosen effect size, and sampled at the repetition time. Noise is autocorrelated, and slow drift is added. A general linear model, with discrete cosine drift regressors when the high-pass filter is on, is fitted by ordinary least squares. Power is estimated by simulating many group studies, in which each participant’s estimate varies with the measurement error of the design and between participants. Head motion and physiological noise are not simulated.
- Friston, K. J., Fletcher, P., Josephs, O., Holmes, A., Rugg, M. D., & Turner, R. (1998). Event-related fMRI: characterizing differential responses. NeuroImage, 7(1), 30–40. https://doi.org/10.1006/nimg.1997.0306
- Dale, A. M. (1999). Optimal experimental design for event-related fMRI. Human Brain Mapping, 8(2–3), 109–114. https://doi.org/10.1002/(SICI)1097-0193(1999)8:2/3<109::AID-HBM7>3.0.CO;2-W
About the simulations
Each simulator is a simplified version of a published model of cells, signals, instruments or behaviour, and runs in the web browser. The simulations show how the techniques work and how design choices change the results; they do not replace training in a laboratory.
All data are simulated and generated afresh on each run, so results vary between runs, as in real experiments. Nothing is stored or sent: designs and results are lost when the page is closed, unless the protocol is copied or the data file is downloaded.
Last updated