Sunday, June 29, 2008

All about fMRI

I am working in a lab that is using fMRI brain scans, with the intent to figure out how the brain makes decisions. An fMRI machine is typically used to detect changes in blood oxygen. There's some major magnetic difference between oxygenated hemoglobin and deoxygenated hemoglobin, and the scans taken are called BOLD (Blood Oxygen Level Dependant) fMRI scans. The theory goes, we know that firing neurons use up oxygen and glucose. So if a certain brain area is using up more oxygen, it's probably more active. And if we can figure out a well-controlled experiment, and find that a certain brain area uses more oxygen during a specific task, well then that brain area may be used to make the computations for that task.

Pure Pedantry over at Scienceblogs explains why fMRI isn't the Grail of neuroscience. And I have to agree. However, Jake's complaints take on a kind of straw-man-ish feel to them. I mean, he mainly just picks on what he calls (hilariously, might I add) OMGFMRI stories. "My amygdala's lighting up, so I must be afraid!" That kind of thing.

I'd like to point out that the paper that Jake linked to has a lot more high quality nitpicks about the use of fMRI, and I'd highly recommend reading that paper.

For example: fMRI may tell you that neurons in an area are active, but it can't tell you what they're doing. Those active neurons might be excitatory neurons, sure. They might be inhibitory. They might be memory-related. They might just be responding to an itch that the participant had. Or they might be a tangental, modulatory process.

Or: fMRI shows a stronger 'signal' when large numbers of neurons in an area are responding. If a process only uses a small number of neurons in any one area, then the effect of the stimulus might be drowned out by attention, memory, and other related (but not relevant) processes. Making better fMRI scanners won't help this.

Basically, the brain is an eight-pound processor. It's massively complicated. It does a lot of shit. It, in fact, controls everything we do. We don't even know why we do everything we do.

On top of that, fMRI is a very broad tool. It's like trying to figure out what your computer is doing by looking at how much electricity is flowing to each component part. Well, your video card is flipping a lot of bits; maybe you're playing a graphics-intensive game. Your hard drive is grinding like mad; maybe you're installing something. Maybe. It's hard to tell.

Oh, and don't even get me started on the whole astrocytes thing.

So is fMRI useless? I'll quote wikipedia on this one: "Like any other technique, fMRI is as worthwhile as the design of the experiment using it." My lab seems to have gotten some decent results, and the procedure makes sense to me (albeit, I am merely an undergrad).

Now that I've summed that up, my head is spinning. I shall go lie down now. Carry on!


  1. clarity++;

    Keep up the good work!

  2. The BOLD fMRI signal observed in brain activation is from a measurement of the relative quantities of oxy and deoxyhemoglobin. Vasodilation increases blood flow in the activated regions and that changed oxy/deoxy ratio is what is observed. Vasodilation is controlled by NO.

    The regions of activation observed in BOLD fMRI are actually regions of NO, where the prompt neurogenic NO release is high enough to cause vasodilation by activating sGC. That NO does things besides vasodilation. Those things are not understood. I think that those things are actually more important than the increased O2 consumption.

    It is well established that the O2 delivery by the increased blood flow exceeds the metabolic requirements of the activated region.