Frequently Asked Questions - Physiology and fMRI
1. Why would I want to collect physiological data?
No, the real question should be: why wouldn't you want to collect physiological data? And the only answer is: because you hate freedom. Haha! Phew.
Actually, the main reason is because physiological effects can be a significant source of noise in your data. The pulsing of blood vessels with the cardiac cycle can move brain tissue, jostle ventricles and alter BOLD signal in specific regions; respiration can induce magnetic inhomogeneities and create head movements. If you measure physiology, you can, at least in part, account for those sources of noise and remove their confounding effects from your data, boosting your signal to noise ratio.
Secondarily (well, primarily, for some), for many studies, physiological measures like heart rate or respiration rate can be an important source of data themselves, providing an important test of autonomic arousal that may supplement self-report data or other measures. All those uses are beyond the scope of this discussion, but it's something to keep in mind.
2. How do I do it?
Thankfully, scanner designers have generally had the foresight to think someone might want to collect this info, and so most scanners have instruments built in to record at least two main physiological measures - heartrate/cardiac cycle (usually with a photoplethysmograph - a small clip that goes on the finger) and respiration (usually with a pneumatic belt - a thin belt that wraps around the chest). For scanners without these instruments built in, several companies manufacture MR-compatible versions of these instruments.
There may be other measures that you may want to collect - galvanic skin response, for example - and here at Stanford, those sensors are relatively easily available. Other measures have less of an effect on creating noise in the signal, however, and so we'll primarily address those two measures.
3. How might the cardiac cycle influence my signal?
Several ways. The pulsation of vessels creates a variety of other movements, as CSF pulses along with it to make room for incoming blood, tissue moves aside slightly as vessels swell and shrink, and waves of blood (and the accompanying BOLD signal) move through the head. This sounds like the effects can be relatively small, but large structures of the brain can move significant amounts in the neighborhood of large vessels, and the resulting motion can significantly change your signal. Generally, the term "pulsatility" is used to describe the process that generates artifacts through cardiac movement. As well, because TRs for many experiments are slower than the cardiac cycle, these effects can occur image-to-image in an unpredictable way, as various points in the cycle are sampled in an irregular fashion.
These effects are more significant in some regions of the brain than in others; Dagli et. al (PhysiologyPapers) address the question of where pulsatility artifact is worst. Perhaps unsurprisingly, areas near large vessels tend to rank among those regions, but other areas are also affected - regions near the borders of are also significantly affected.
4. How might respiration influence my signal?
Also a couple ways. Respiration, by its nature, can cause the head and particular parts of it (sinus cavities, etc.) to move slightly, which can induce motion-related changes in signal. Perhaps more significantly, the inflating and deflating of the lungs changes the magnetic signature of the human body, and that signature change can induce inhomogeneities in the baseline magnetic field (B0) that you've carefully tuned with your shim. Those inhomogeneities can be unpredictable and can affect your signal in unpredictable ways. Breathing rate can also be significantly less predictable than cardiac cycle - many subjects take spontaneous deeper breaths at irregular intervals, for example. Van de Moortele et. al (PhysiologyPapers) address the sources of respiration artifact in some detail.
5. What can I do to account for these changes?
Thought you'd never ask. There are several ways. Pfeuffer et. al (PhysiologyPapers) present a navigator-based method in k-space that adjusts, in large part, for global effects, more due to respiration changes. Perhaps the most prevalent ways, though, are retrospective, and rest on the fact that generally, cardiac cycle and respiration cycle take place at a time scale far faster than stimulus presentation for most experiments. Isolating signal changes that take place at the appropriate frequency, then, can help isolate those sources of noise.
This isn't as easy as it sounds, due to the potential aliasing of this noise because of the difference between TR and cardiac cycle time. But it is possible, with some work. Glover et. al (PhysiologyPapers) present one of the industry-standard ways of doing this correction - by sorting images according to their point in the cardiac or respiration cycle, the appropriate amount of signal change due to those sources can be identified and removed from the image. This correction happens at the point of reconstruction of images from raw data, and is available automatically at Stanford in the makevols program.
Care should be taken with this option, though. As with any type of algorithm that removes "confounding" signals (realignment, say), this correction can't account for the extent to which physiological noise is correlated with the task. If there is significant correlation between your task and your physiological measures, removing noise due to the physiological sources will also remove task-related signal. This could happen with rapid event-related tasks if subjects breath in time with the tasks, or in any sort of experiment that might induce arousal - emotionally arousing stimuli may increase heart rate and respiration rate and thus change those noise profiles in a task-correlated way. In this case, Glover et. al suggest using only resting-state images to calculate this correction; this is always a significant consideration in physiological artifacts.
6. Are there other sources of physiological noise I might want to worry about?
Possibly. Peeters and Van der Linden (PhysiologyPapers) address the question of longer-term physiological changes, such as those induced by pharmacological manipulations or sudden environmental changes. As a drug begins to be absorbed, for example, there can be a gradual change in vasoconstriction or blood oxygenation in general that can look like a global signal drift but is, in fact, a changing of the sources of the signal. Researchers interested in looking at these sorts of changes should look with care at these sorts of corrections.