fMRI

fMRI is a special application of MRI used to look at brain function, more or less. As the various parts of the brain work, they demand freshly oxygenated blood; the sudden perfusion of blood to a particular brain area in response to increased workload causes the magnetic susceptibility of that brain area to ever so slightly change -- In other words, the signal emitted by any particular piece of brain will fluctuate over time depending on its supply of oxygenated blood and the rate at which it metabolilizes that supply.

The cool part is this: if you know what your particular brain is thinking about at any given time during the scan (i.e., you give the brain some interesting cognitive task to do,) you can associate brain the signal changes you see with the task that the brain is working on.

The most common application of fMRI is in research, usually cognitive neuroscience research. Psychiatrists and radiologists also use fMRI methods in the hopes that they may be some day developed to the point they can be used to diagnose or even treat disesase.

And the BEST part is that fMRI requires *really* powerful magnets, fiber optic button boxes and ton's of NIH dollars.

fMRI Paradigm

The underlying assumption behind all fMRI research is the idea that a working brain requires more oxygen than a resting brain, and a brain performing a difficult task would require more oxygen still. Essentially:

Increase in Information Processing
leads to
Increase in Neural Activity
leads to
Increase in Blood Flow
leads to
Increase in Venous Oxygenation

It is this venous oxygenation that fMRI machines register and we interpret as representing neural activity. Because it's technically blood flow that is being directly measured and not thought, certain steps must be taken to isolate the specific areas of activation as much as possible while filtering out the incidental properties of blood flow.

• For one thing, the veins and arteries that feed the brain must be subtracted from the data because the blood in them responds to the magnet just as well as the blood being used by the brain itself. If the veins were not subtracted, the data that registers the activity would be clouded by the normal areas of blood flow--not very good for pinpointing areas of neural activity.
• Secondly, because we are relying on blood flow to the brain, the data obtained from the fMRI is a bit "sloshy", meaning that it does not have good temporal resolution. When a particular stimulus is presented and the brain starts working harder to accomplish the task, it takes a while for the increase in blood flow to occur, sometimes in extreme cases up to six seconds. Because of this, experiments are run in "block trials" where the same sort of task is presented many times in a row. This allows the blood enough time to reach the brain so that the fMRI machine can track it.

While being very good in terms of locating activity, block trials for obvious reasons are very poor at any time-dependant tasks. If one needs to study such things, ERP (event related potential) is a better measure.

Other MRI Info

• Most MRIs come in two strengths - 1.5 Tesla and 4 Tesla (very strong magnetic field)*
• Research MRI machines can cost up to $300 an hour to use (necessitating big grants)
• When the machine is in use, it makes a lot of very loud noises, sometimes sounding like clicks, sometimes more like a beeping. The sound is actually generated by the magnet as it torques to shift the magnetic field.
• Even though the side effects are unknown, pregnant women are prohibited by law from being part of any fMRI research.

---

*Actually, the magnets come in .5, 1.5, 3, 4, and 7 Tesla strengths. (thanks to qeyser for the info!)