Head Movement Is Bad News For Neuroscience (Again)

By Neuroskeptic | November 26, 2013 3:53 pm

Claims that children with autism have abnormal brain white matter connections may just reflect the fact that they move about more during their MRI scans.

So say a team of Harvard and MIT neuroscientists, including Nancy “Voodoo Correlations” Kanwisher, in a new paper: Spurious group differences due to head motion in a diffusion MRI study.

Essentially, the authors show how head movement during a diffusion tensor imaging (DTI) scan causes apparant differences in the integrity of white matter tracts, like these ones:


In comparisons of two randomized groups of healthy children – in whom no white matter differences ought to appear – spurious effects were seen whenever one group moved more than the other:


As for autism, the authors found that kids with autism moved more, on average, than controls, and that matching the two groups by motion reduced the magnitude of the group differences in white matter (though many remained significant).

Technically, the motion-related differences manifested as increases in RD and reductions in FA; these were localized:

The pathways that exhibited the most substantial motion-induced group differences in our data were the corpus callosum and the cingulum bundle. Perhaps this is related to their proximity to non-brain voxels (such as the ventricles) … deeper brain areas appear to be more affected than more superficial ones, thus distance from the head coils may also be a factor.

The good news is that there’s a simple fix: entering the motion parameters, extracted from the DTI data itself, as a covariate in the analysis. The authors show that this is extremely effective. The bad news is that most researchers don’t do this.

Attentive readers might remember that the motion issue has already been raised in an autism context in regards to functional connectivity (in fact I believe that I was the first person to put this concern in writing.) This time it’s about structural connections.

But autism isn’t the only disorder where this might be an issue. My working assumption is that every ‘disease’ group moves more than every ‘healthy’ group – until proven otherwise.

ResearchBlogging.orgYendiki A, Koldewyn K, Kakunoori S, Kanwisher N, & Fischl B (2013). Spurious group differences due to head motion in a diffusion MRI study Neuroimage DOI: 10.1016/j.neuroimage.2013.11.027

  • http://psychiatrietogo.wordpress.com/ jan.dreher@me.com

    Every desease group moves more, less or in a different pattern than the healthy control group! There is a lot to measure in an fMRI, it will be interesting to see, what exactly was measured. But this could take some more years to see!

  • Pingback: Head Movement Is Bad News For Neuroscience (Aga...()

  • http://petrossa.me/ petrossa

    Sigh. fMRI … When will it die a silent death finally.

    • Defenestrator

      This article is not even about fMRI.

      • http://petrossa.me/ petrossa

        I always tend to overestimate the comprehension level of readers :( If even MRI doesn’t work properly how could possibly fMRI work? I hope this made it more understandable for you.

        • Richard

          Structural MRI is fine (at a given resolution), but the DTI method is looking at fine fibre tracts, hence the large movement issue. Your argument doesn’t follow…

          • http://petrossa.me/ petrossa

            fMRI is also very dependent on head movement. Argument follows easily

          • Richard

            It is, but fMRI is measuring something different from structural MRI & DTI. Movement isn’t necessarily going to affect each technique similarly.

          • http://petrossa.me/ petrossa

            fMRI is even worse then MRI since it depends on separating out tiny signal from lots of noise, introducing even more noise due to movement makes it as detailed as an old style EEG with only a few electrodes. You can determine with very low accuracy at the best of times, and even then its not totally sure you didn’t set your thresholds correctly. How are going to know? So seeing a blob light up, ok. Deciding if a tiny part X takes part in higher order process Y NO WAY. Pisses me off to see the most absurd papers based on fMRI with claims about having found the common part that connects religion and logic, or how a female orgasm opens the consciousness. Those kind of ridiculous papers got published in various journals. The technique should be shelved and be replaced by something more modern. This is a waste of money, time and my bloodpressure control meds.

          • Richard

            Right, but as Defenestrator noted, that doesn’t have much to do with this article… Do you think structural MRI and DTI are inherently useless as well?

          • practiCalfMRI

            Some technical clarification for those interested in the topic. (I’m leaving the arguments to you guys!)

            Diffusion-weighted MRI, which is the parent acquisition method used to acquire DTI and the other tractographic methods, is intrinsically motion-sensitive, by design. The whole idea is to make the acquisition sensitive to motion on the length scale of water diffusion at 37 C. Sadly, it also makes the acquisition sensitive to motions greater than that, too. If the bulk subject motion is isotropic then that leads to diffusion maps with low signal-to-noise. If the bulk subject motion is anisotropic, as it most often is, then that will lead to some sort of systematic bias in the final data. There are tricks to try to reduce the anisotropy of data quality, e.g. re-acquire bum component images, but you can only fix that which you know (or suspect) to be broken; bias can and likely will remain at some level.

            As for fMRI, it is not designed to be sensitive to motion, it just happens to be so (along with all other MRI methods). We can consider fMRI to be more motion-sensitive than anatomical MRI scans on three grounds, however: (1) the duration of the EPI time series is typically longer than for an anatomical scan; (2) the individual fMRI images use a contrast (BOLD) that will accentuate the effects of very small head movements; (3) the images are analyzed as a time series so anything that changes during the series will be interpreted as some sort of “functional” change, even if it’s junk.

          • Defenestrator

            Exactly. And the BOLD response effects of motion are almost entirely at the borders of the brain and are pretty easy to pick out. Not to mention, you can look for motion and simply regress it, which is harder to do with DTI scans.

          • http://petrossa.me/ petrossa

            wow. jargon. We’re getting serious. Let me get my dictionary. Listen it’s very very simple no need for big words to show off your immense knowledge. When you have a very low signal to noise ratio it’s hard enough to filter out the signal. The last thing you need is additional noise. That coupled with you don’t know what you’re looking for and where since you have no idea where to put your masks (or you just mask off stuff that is in fact related but it doesn’t fit your presupposed position) there is a huge room for error.
            That coupled with the fact that mere oxygenation doesn’t tell anything but something is active, not what it’s doing you might as well just draw the image using photoshop in cases where you claim to ‘discover’ some high resolution area.
            Smug condescension never made for a good argument.

          • Mikey Godsey

            I just read the entire exchange. You were simply wrong, because you don’t know enough about MRI or fMRI technology. You got an attitude first.

            You were wrong, and you know what? THAT’S OKAY. Calm down and read a book before you tell someone they’re wrong about something they clearly have vastly more knowledge about than you do.

          • http://petrossa.me/ petrossa

            funny. Must be some kind of blind spot in commentators. The bland assumption i didn’t study the matter profoundly before stating my opinion just because it doesn’t conform to your opinion. It doesn’t get more condescending then that. You know, the ‘you are stupid because authority and so they know’ argument doesn’t work. I seriously doubt that those who publicize the more absurd papers, which claim to be able to determine to a higher degree of resolution then an EEG to find definitive ‘spots’ which do this or that even know how their toy works. Which actually i do, having been a systems analyst/programmer involved in writing statistical models. Yep, now it’s my turn to draw the ‘i am an authority’ card. Equal footing

          • Mikey Godsey

            If you know how it works why are you so wrong?

          • http://petrossa.me/ petrossa

            disagreeing with mainstream thinking doesn’t automatically imply one is wrong. It can also mean one has more objective view on the matter and see flaws not seen by mainstream. At least the writer of this blog has him/herself doubts about the system. Recently he/she posted a tweet to a link of neuro’s indicating some of my objections are getting attention in the community. I guess i’m just picking the signal.

          • Richard

            It’s not the criticism of fMRI that’s wrong, it’s the link between fMRI movement artefacts and DTI movement artefacts you’ve implied. Both are caused by head movements yes, but the effects and implications are different for each modality, so to link problems in fMRI to problems in DTI doesn’t really make sense.

          • http://petrossa.me/ petrossa

            Title of post: “Head Movement Is Bad News For Neuroscience” goes on to mention MRI. So i added fMRI since it is already inherently very iffy, adding head movement to that only makes it even worse. It was not a direct 1 – 1 linkage of effects. Imho the post could have better been written adding a paragraph on the subject. Which was about the original intent of my OP.

          • Mikey Godsey

            You’re right, but replying to someone stating that you’re referring to something other than what was mentioned to the article with “I always tend to overestimate the comprehension level of readers ” was petty, and honestly gave me the impression that you have no idea what you’re talking about. And then later on, “Wow. Jargon”. Seriously? The way you’ve approached this entire discussion makes me think it’s probably a good thing I’ve never met you.

          • http://petrossa.me/ petrossa

            way to go, ad hominem. Discussion over

          • Mikey Godsey

            Try checking out the other reply before you go crying about irrelevance.

          • Mikey Godsey

            Isn’t referring to the comprehension level of the other people involved in the discussion considered ad hominem? Oh wait, you’re right all the time regardless, I forgot. Sorry.

          • Mikey Godsey

            And more more thing: That would be “responding to tone”, not “ad hominem”. Paul Graham’s Hierarchy of Disagreement. I’ll let it go now though, this isn’t the first time you’ve spoke about something and been wrong.

          • Mikey Godsey

            And to be honest, I didn’t say you were wrong because “authority”. I said you were wrong because evidence was laid out by the other commenters (Not commentators. Again, read a book) that showed me you were incorrect. Just because people disagree with you doesn’t mean we’re part of some entrenched power structure.

        • Defenestrator

          Maybe I could try to help you understand the fact that fractional anisotropy calculations have very little to do with how BOLD responses are calculated, and that fractional anisotropy would be much more effected by movement than BOLD response. But I always tend to underestimate the smug ignorance of people on the internet.

  • Pingback: Head Movement Is Bad News For Neuroscience (Aga...()

  • Jared Blackburn

    Artifact is often a serious issue in studies of live human subjects. Its a major issue in quantitative EEG based studies as well, where EMG contamination is often an issue. Anytime you can’t either eliminate or somehow control for movement during an anatomical or physiological of the brain you can expect some misleading results — and its often not easy to do.

  • Pingback: Braincast 332 – I have a lucid dream › braincast › SciLogs - Wissenschaftsblogs()

  • practiCalfMRI

    “But autism isn’t the only disorder where this might be an issue. My
    working assumption is that every ‘disease’ group moves more than every ‘healthy’ group – until proven otherwise.”

    Or between men and women, say. This paper is making headlines all over the place today: http://www.pnas.org/content/early/2013/11/27/1316909110

    For all we know the results may be accurate. But there is no mention of motion as a possible driver of the group differences, in spite of prior demonstrations that gender can drive functional connectivity differences in resting state fMRI: http://www.sciencedirect.com/science/article/pii/S1053811911008214

    Not even a mention of head motion or subject movement, when diffusion imaging is intentionally motion-sensitive. That seems like an oversight to me. Perhaps the authors will re-analyze their data using the covariates of motion approach advocated by Yendiki et al. Until then it’s hard to put too much stock into the results.

  • Pingback: Men, Women, and Big PNAS Papers - Neuroskeptic | DiscoverMagazine.com()

  • Pingback: Brains: sex, society and #neurotrash - ScienceGrrl()

  • feloniousgrammar

    Perhaps, someday, there will be imaging and assessment of images to demonstrate the processes used in derailment.

  • Pingback: Baby Brain Scans Predict Later Cognitive Development? - Neuroskeptic | DiscoverMagazine.com()

  • http://www.conxz.net/blog/aboutme/ Conxz

    Hi, what do you think about the motion effects in structural MRI data?

    • http://blogs.discovermagazine.com/neuroskeptic/ Neuroskeptic

      I don’t know much about that personally but from what I hear, motion can cause many artifacts, however some structural sequences are more robust to them than others.

  • Pingback: fMRI Motion Correction: The Quick and the Dead - Neuroskeptic | DiscoverMagazine.com()

  • DS

    “The good news is that there’s a simple fix ..”

    It is difficult to publish such a paper without claiming a fix. How many motion “fixes” have we had with respect fMRI? This too shall not pass.

  • Pingback: The Inherent Limits of MRI Tractography? - Neuroskeptic | DiscoverMagazine.com()

  • Pingback: The Inherent Limits of MRI Tractography? | AllDigitalNews.com()

  • Pingback: The Inherent Limits of MRI Tractography?Fresh News Today | Fresh News Today()

  • Pingback: Head Motion Biases Brain Structural Scans - Neuroskeptic()

  • Pingback: Head Motion Biases Brain Structural Scans * The New World()

  • Pingback: Head Motion Biases Brain Structural ScansFresh News Today | Fresh News Today()



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