fMRI: Adrift on Ten-Second Waves?

By Neuroskeptic | November 6, 2013 5:10 pm

For the first time, neuroscientists have directly observed a slow, steady fluctuation – a ‘wave’ – in the blood flow to the brain.

The oscillation, which has a frequency of 0.1 Hz, or one cycle every 10 seconds, is mysterious, and could have big implications for neuroscience.

The researchers, led by Aleksandr Rayshubskiy of Columbia university, used a fancy camera to record high-resolution movies of the surface of the brain of a patient who’d had her skull opened up during surgery for a brain tumour.

By illuminating the brain with red, green, and blue lights, Rayshubskiy et al were able to measure oxygenated (HbO) and deoxygenated (HbR) blood along with total blood flow (HbT) – all thanks to the fact that oxygenated blood is redder.

This revealed that in some bits of the brain surface, the amount of total blood, and oxygenated blood, changed with a consistent 0.1 Hz rhythm:


A second patient, with epilepsy, did not show the same activity, however.

Further investigation showed the waves to be closely correlated with the expansion and contraction of certain small arteries. Therefore, the phenomenon seems to be myogenic (muscle-based) in origin, as it’s smooth muscle that controls artery size.

vesselThis isn’t the first time that 10-second blood flow waves have been suggested to exist in the brain, but until now, they’ve never been observed directly – and their origin has been a mystery. We now know it’s myogenic although what function it serves is anyone’s guess.

But why does it matter?

As a neuroscientist, I’m most interested in what it could mean for fMRI scans. fMRI is a cornerstone of modern neuroscience and it’s based on measuring the amount of deoxygenated haemoglobin (HbR) in different brain regions.

fMRI picks up neural activity indirectly because brain activity uses oxygen, and therefore alters HbR. Now the changes that fMRI picks up last around 6-12 seconds… in other words, a neural activation “peak” could easily be confused with one of those 0.1 Hz waves.

If that’s not bad enough, these waves would be especially harmful for functional connectivity fMRI. If these 10-second waves extend over large chunks of the brain, they might look very much like a coherent pattern of activation. Nightmare.

However… let’s not panic yet.

The 0.1 Hz cycle that Rayshubskiy found, remember, did not involve deoxyhaemoglobin at all – it was a change in oxygenated blood. Which makes sense, because that’s the kind of blood that arteries carry.

So one wouldn’t expect fMRI to be affected, at least not directly. Phew!

However… actually, Rayshubskiy et al claim that 10-second cycles do show up on fMRI (though they don’t try to explain it.) Here’s their fMRI data from the same patient who got the brain camera:

fmriThis data was recorded using an standard fMRI sequence, with a time resolution of 2 seconds. A 0.1 Hz wave is seen in the areas they show, but I do worry that they might have selected those areas precisely because they show that.

They say

…manual inspection of voxel time-courses [was used] to identify the ROIs and corresponding time-courses shown in Figures 4D, E and F.

Which sounds a lot like the old fMRI ‘voodoo error‘ i.e. cherry picking brain areas that fit your model (perhaps by chance).

So what I’m saying is, this fMRI result might be voodoo… and that would be good news for fMRI. Rather ironic.

Luckily, it should be easy to try to replicate these results, because the Rayshubskiy fMRI protocol was a standard one. There are hundreds of neuroscientists who could reanalyze some comparable existing data, and look for 0.1 Hz oscillations.

If they don’t show up, the 10 second waves would still be an interesting physiological phenomenon, but one that fMRI users might be able to safely ignore.
ResearchBlogging.orgRayshubskiy A, Wojtasiewicz TJ, Mikell CB, Bouchard MB, Timerman D, Youngerman BE, McGovern RA, Otten ML, Canoll PD, McKhann GM 2nd, & Hillman EM (2013). Direct, intraoperative observation of approx. 0.1 Hz hemodynamic oscillations in awake human cortex: implications for fMRI. NeuroImage PMID: 24185013

CATEGORIZED UNDER: fMRI, methods, papers, select, Top Posts, voodoo
  • DS

    Hmm. Spin history effects last for approximately 10 seconds.

  • K Jinx

    The phenomenon beeing myogenic does not rule out a neural cause of the muscular activity. In fact, the BOLD-signal is strongly dependent on small artery muscles because of feedback mechanisms between neural and muscular activity. So the 0.1 Hz oscillations might actually reflect neural resting state activity. Allthough most studies I know have found coherent neural resting state oscillations at a rate somewhat lower than 0.1 Hz.

    • practiCalfMRI

      Exactly. If the brain has evolved certain ongoing synchronous activity patterns then supporting synchronous blood changes would make sense. A simultaneous fNIRS-ECog study would be interesting.

      As for BOLD, from their graph it would seem that HbT would drive the possible signal change; HbR changes are small. I would first be interested to see replication in the presence of head motion, a problem that is bigger for BOLD than optical signals. I’m not yet convinced it will be easy to detect in fMRI, unless it is the same origin as resting state changes after all.

    • fMRI_today

      Completely agree as well

  • fMRI_today

    When you say “directly observed,” do you mean visually? I’m not sure if that’s more “real” than other measures. My understanding is that low frequency oscillations in flow were characterized well in the early 90’s in animals. Hudetz et al is an example. By 2000, these oscillations in flow were characterized well in humans. Obrig et al is an example. Also, this article is confusing in that the observation of Rayshubskiy may be in fact the useful connectivity-related oscillations that we all have come to love. Increase in arterial flow directly leads to BOLD changes. Since he didn’t measure the whole brain there is no proving that these oscillations are unique from fMRI BOLD connectivity related oscillations.

    • Neuroskeptic

      I agree, but isn’t that the worry? That fMRI BOLD connectivity might be driven by (and an epiphenomenon of) an essentially myogenic process – as opposed to being products of brain activity in connected regions?

      • practiCalfMRI

        But BOLD is already a proxy for neural activity, loosely defined as collective electrical signaling. The connectivity issue isn’t addressed by this study, as Peter intimates, we need more (optical) data from a larger patch of brain. If this stuff is real then it shouldn’t be long before others report such studies. Presumably a scalp fNIRS study should find these signals if they are robust. Blaise Frederick or someone else may already have seen them, I’ve not looked at the pertinent literature. Will take a look when I’m back from travel.

      • fMRI_today

        That would be a worry if BOLD resting state connectivity were not so well established as reflecting functional architecture. Many papers have been published showing the exquisite correspondence of resting state connectivity to functional activation locations. So if resting state reflects something other than ongoing neuronal processes, whatever it is based on sure seems to spatially correspond to functional organization of the brain. Many unknowns!

  • DS

    Why did the authors use slice-timing correction? I see no reason for having done so. Furthermore, since slice-timing correction and realignment should not in principle be separated (and do not commute) then even their minimal processing steps draw the results into question.

  • corticalcolumns

    There’s already prior evidence for 0.1 Hz oscillations in the
    brain and circulation system (e.g. Mayer waves). The question is what does this
    reflect? What the hell would be so stringently periodic on this timescale? Too
    slow for neural stuff, too periodic for behavior probably not related to slow
    rhythms like circadian stuff – so what? If there’s no answer to that then this
    is not particularly interesting…

  • Pingback: Neuroscience 2013 – Nachklapp › braincast › SciLogs - Wissenschaftsblogs()



No brain. No gain.

About Neuroskeptic

Neuroskeptic is a British neuroscientist who takes a skeptical look at his own field, and beyond. His blog offers a look at the latest developments in neuroscience, psychiatry and psychology through a critical lens.


See More

@Neuro_Skeptic on Twitter


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Collapse bottom bar