Insula and Anterior Cingulate: the ‘everything’ network or systemic neurovascular confound?

It’s no secret in cognitive neuroscience that some brain regions garner more attention than others. Particularly in fMRI research, we’re all too familiar with certain regions that seem to pop up in study after study, regardless of experimental paradigm. When it comes to areas like the anterior cingulate cortex (ACC) and insula (AIC), the trend is obvious. Generally when I see the same brain region involved in a wide a variety of tasks, I think there must be some very general level function which encompasses these paradigms. Off the top of my head, the ACC and AIC are major players in cognitive control, pain, emotion, consciousness, salience, working memory, decision making, and interoception to name a few. Maybe on a bad day I’ll look at a list like that and think, well localization is just all wrong, and really what we have is a big fat prefrontal cortex doing everything in conjunction. A paper published yesterday in Cerebral Cortex took my breath away and lead to a third, more sinister option: a serious methodological confound in a large majority of published fMRI papers.

Neurovascular coupling and the BOLD signal: a match not made in heaven

An important line of research in neuroimaging focuses on noise in fMRI signals. The essential problem of fMRI is that, while it provides decent spatial resolution, the data is acquired slowly and indirectly via the blood-oxygenation level dependent (BOLD) signal. The BOLD signal is messy, slow, and extremely complex in its origins. Although we typically assume increasing BOLD signal equals greater neural activity, the details of just what kind of activity (e.g. excitatory vs inhibitory, post-synaptic vs local field) are murky at best. Advancements in multi-modal and optogenetic imaging hold a great deal of promise regarding the signal’s true nature, but sadly we are currently at a “best guess” level of understanding. This weakness means that without careful experimental design, it can be difficult to rule out non-neural contributors to our fMRI signal. Setting aside the worry about what neural activity IS measured by BOLD signal, there is still the very real threat of non-neural sources like respiration and cardiovascular function confounding the final result. This is a whole field of research in itself, and is far too complex to summarize here in its entirety. The basic issue is quite simple though.

End-tidal C02, respiration, and the BOLD Signal

In a nutshell, the BOLD signal is thought to measure downstream changes in cerebral blood-flow (CBF) in response to neural activity. This relationship, between neural firing and blood flow, is called neurovascular coupling and is extremely complex, involving astrocytes and multiple chemical pathways. Additionally, it’s quite slow: typically one observes a 3-5 second delay between stimulation and BOLD response. This creates our first noise-related issue; the time between each ‘slice’ of the brain, or repetition time (TR), must be optimized to detect signals at this frequency. This means we sample from our participant’s brain slowly. Typically we sample every 3-5 seconds and construct our paradigms in ways that respect the natural time lag of the BOLD signal. Stimulate too fast, and the vasculature doesn’t have time to respond. Stimulation frequency also helps prevent our first simple confound: our pulse and respiration rates tend oscillate at slightly slower frequencies (approximately every 10-15 seconds). This is a good thing, and it means that so long as your design is well controlled (i.e. your events are properly staggered and your baseline is well defined) you shouldn’t have to worry too much about confounds. But that’s our first problematic assumption; consider for example when one’s paradigms use long blocks of obscure things like “decide how much you identify with these stimuli”. If cognitive load differs between conditions, or your groups (for example, a PTSD and a control group) react differently to the stimuli, respiration and pulse rates might easily begin to overlap your sampling frequency, confounding the BOLD signal.

But you say, my experiment is well controlled, and there’s no way my groups are breathing THAT differently! Fair enough, but this leads us to our next problem: end tidal CO2. Without getting into the complex physiology, end-tidal CO2 is a by-product of respiration. When you hold your breath, CO2 blood levels rise dramatically. CO2 is a potent vasodilator, meaning it opens blood vessels and increases local blood flow. You’ve probably guessed where I’m going with this: hold your breath in the fMRI and you get massive alterations in the BOLD signal. Your participants don’t even need to match the sampling frequency of the paradigm to confound the BOLD; they simply need to breath at slightly different rates in each group or condition and suddenly your results are full of CO2 driven false positives! This is a serious problem for any kind of unconstrained experimental design, especially those involving poorly conceptualized social tasks or long periods of free activity. Imagine now that certain regions of the brain might respond differently to levels of CO2.

This image is from Change & Glover’s paper, “Relationship between respiration, end-tidal CO2, and BOLD signals in resting-state fMRI”. Here they measure both CO2 and respiration frequency during a standard resting-state scan. The image displays the results of group-level regression of these signals with BOLD. I’ve added circles in blue around the areas that respond the strongest. Without consulting an atlas, we can clearly see that bilateral anterior insula extending upwards into parietal cortex, anterior cingulate, and medial-prefrontal regions are hugely susceptible to respiration and CO2. This is pretty damning for resting-state fMRI, and makes sense given that resting state fluctuations occur at roughly the same rate as respiration. But what about well-controlled event related designs? Might variability in neurovascular coupling cause a similar pattern of response? Here is where Di et al’s paper lends a somewhat terrifying result:

Di et al recently investigated the role of vascular confounds in fMRI by administrating a common digit-symbol substitution task (DSST), a resting state, and a breath-holding paradigm. Signals related to resting-state and breath-holding were then extracted and entered into multiple regression with the DSST-related activations. This allowed Di et al to estimate what brain regions were most influenced by low-frequency fluctuation (ALFF, a common resting state measure) and purely vascular sources (breath-holding). From the figure above, you can see that regions marked with the blue arrow were the most suppressed, meaning the signal explained by the event-related model was significantly correlated with the covariates, and in red where the signal was significantly improved by removal of the covariates. The authors conclude that “(results) indicated that the adjustment tended to suppress activation in regions that were near vessels such as midline cingulate gyrus, bilateral anterior insula, and posterior cerebellum.” It seems that indeed, our old friends the anterior insula and cingulate cortex are extremely susceptible to neurovascular confound.

What does this mean for cognitive neuroscience? For one, it should be clear that even well-controlled fMRI designs can exhibit such confounds. This doesn’t mean we should throw the baby out with the bathwater though; some designs are better than others. Thankfully it’s pretty easy to measure respiration with most scanners, and so it is probably a good idea at minimum to check if one’s experimental conditions do indeed create differential respiration patterns. Further, we need to be especially cautious in cases like meditation or clinical fMRI, where special participant groups may have different baseline respiration rates or stronger parasympathetic responses to stimuli. Sadly, I’m afraid that looking back, these findings greatly limit our conclusions in any design that did not control for these issues. Remember that the insula and ACC are currently cognitive neuroscience’s hottest regions. I’m not even going to get into resting state, where these problems are all magnified 10 fold. I’ll leave you with this image from neuroskeptic, estimating the year’s most popular brain regions:

Are those spikes publication fads, every-task regions, or neurovascular artifacts? You be the judge.

edit:As many of you had questions or comments regarding the best way to deal with respiratory related issues, I spoke briefly with resident noise expert Torben Lund at yesterday’s lab meeting. Removal of respiratory noise is fairly simple, but the real problem is with end-tidal C02. According to Torben, most noise experts agree that regression techniques only partially remove the artifact, and that an unknown amount is left behind even following signal regression. This may be due to slow vascular saturation effects that build up and remain irrespective of shear breath frequency. A very tricky problem indeed, and certainly worth researching.
Note: credit goes to my methods teacher and fMRI noise expert Torben Lund, and CFIN neurobiologist Rasmus Aamand, for introducing and explaining the basis of the C02/respiration issue to me. Rasmus particularly, whose sharp comments lead to my including respiration and pulse measures in my last meditation project.

20 thoughts on “Insula and Anterior Cingulate: the ‘everything’ network or systemic neurovascular confound?

  1. I don’t know enough about fMRI and BOLD methodology to properly comment on the paper you have reviewed here, but I like to contrast the imaging method of brain-behaviour relationships with the lesion method. As you know, imaging methods can at best show an association between activity in a particular brain area, whereas lesion studies can be considered a stronger argument in favour of a particular brain area being critical to *causing* a particular behaviour.

    I know a survivor of acquired brain injury with damage to the anterior cingulate cortex, who is severely affected with akinetic mutism. He apparently lacks any drive or motivation to speak, and moves very slowly and torpidly, despite being able to speak (I have heard him speak once in the last 6 years, although I only see him every few months). In the literature, Damasio describes a patient who recovered from akinetic mutism, who describes the experience as being totally devoid of any desire do say or do anything (Damasio & Van Hoesen, 1983).

    I greatly admire imaging research, and the new possibilities these methods have given us for understanding brain-behaviour relationships, but sometimes feel that the lesion method is ignored by the media, as the less sexy sister of imaging research. Perhaps this is because work with survivors of brain injury can be a more depressing story, or perhaps it is because imaging methods use fancy machines and produce media friendly images. However, I think that to truly gain perspective on causal involvement of brain areas on cognition and behaviour, we need to consider both methods.

    Perhaps you could write a bit about the lesion method here? (apologies if you already have, I am relatively new to your blog). A great review paper on the lesion method is one by Van Orden (2001) entitled: What do double dissociations prove? Its quite critical of the lesion method too, which is important.


    Damasio, A. R., & Van Hoesen, G. W. (1983). Emotional disturbances associated with focal lesions of the limbic frontal lobe. In K. M. Heilman & P. Satz (Eds.), Neuropsychology of Human emotion. New York: The Guildford Press.

    Van Orden, G. C., Pennington, B. F., & Stone, G. O. (2001). What do double dissociations prove? Cognitive Science, 25, 111-172.

    • Thanks for your interesting comment. Funny that you reference that van Orden paper, the thing literally just came across my desk two days ago. It looks very interesting and I hope to get around to reading it soon. My quick response to you is that, while there are (arguable) explanatory benefits to the lesion method over pure correlation, the lesion method is itself problematic in many ways. The two main problems, I’m sure your aware, are that lesions are rarely focal in nature. Even in the case of a focal lesion, the fact that one could only ever hope to find a handful of individuals limits the generalizability of the findings. Particularly if you, like I do, believe that the mature brain is defined by radical individual differences in both structure and function.

  2. This is interesting, but one of the problems with your dire interpretation of the paper is that we have no idea of the magnitude of this “suppression”. If the t-stat goes from 5 to 4.5 after adjustment, this won’t change the probability a whole lot and the conclusions stay the same. Reading the results of the cerebral cortex paper doesn’t help as all the descriptions are qualitative (this region was suppressed, this region was enhanced, no statistics!). This is largely ok for the point they authors are trying to make, but here you’ are arguing that fMRI based cognitive neuroscience is in big trouble (i.e. “paper published yesterday in Cerebral Cortex took my breath away and lead to a third, more sinister option: a serious methodological confound in a large majority of published fMRI papers.”)

    But really it’s not. If you look at figures 7 and 9, you can see that the overlap between the original and adjusted activation maps is very very high, even in the anterior insula and ACC. Therefore if you ran this study and adjusted (or did not adjust) for neurovascular confounds, your interpretation wouldn’t change in the slightest bit. It’s surely a good thing to adjust for all sources of error that you can think of, but in this case the paper doesn’t convince me that this adjustment is of critical importance for all future fMRI work. Nor do I think the authors of the paper are trying to say that it is. They present it as a new, and seemingly easy to employ, method to remove some confounds and increase specificity. But they never say that not doing so would invalidate anyone’s interpretations to any great extent.

    • Thanks for the great comment vw. I’ve edited some of my overly strong rhetoric regarding confounds in event related designs. You are exactly right that in the case of the DSST, the insula finding is still largely there after correction. Still, the authors estimate that 11.22% of the DSST variance can be attributed to the breath holding + gray matter covariate. I think my point wasn’t that all well-controlled designs will be confounded, but that even in a well controlled design we can see a significant contribution of neurovascular confound. One thing I noted just now looking over the paper is that the authors report a striking pattern of whole brain correlation between ALFF and breath holding. There is a lesser but still telling pattern between the DSST and BH. The take home is that, this is something we can account for, but a great majority of designs fail to do so. It’s relatively harmless in a within-subjects task with a good baseline, but the problem amplifies immensely when we move to between groups or resting baselines. And those two categories probably make up a large majority of current fMRI research.

  3. Indeed, it can’t hurt. I have to say though, your discussion of respiration really did scare me a bit. It’s something we always hear about and figure we should account for, some day, when there’s time. But I think I want to start collecting it now. If only to, as you suggest, compare conditions and ensure respiration is not a confound. And if we can squeeze more power out of a study by correcting for respiration, that’s always welcome.

  4. My inverted logic: If you can measure it during an fMRI experiment without significantly altering the experiment, measure it! I think this applies to both heart rate (where variation also causes issues – see Chang & Glover) and breathing. (BIOPAC will sell you an awesome system for about $10K, or <20 hrs of scan time on most scanners…) To me the onus is on people to show why they *didn't* measure HR/resp.

    And finally:

    It won't make you depressed, but it will make you even more cautious!

    • Thanks, and I agree! In my last study we collected both respiration and pulse using equipment that came with our Seimens Trio. It was very easy and gave me peace of mind knowing that I could compare breath fluctuation and regress it out of my data if needed. I think some of the negative reaction to my post is that people are jumping to defend strong-within subjects design. Again, I agree totally. My own work is with groups (for example meditation) and I wanted to drive home the point that I’m quite worried datasets like mine may be plagued by baseline breathing artifacts. It’s a pretty simple worry if you think about it. One group trains breath awareness and control during challenge, the other doesn’t. Then you challenge that group in fMRI and they alter breathing rates during that particular condition, shifting to a more meditative frequency. Now you’ll be sure to see lots of ACC and Insula showing up in that condition- but it’s not the neural correlates of meditation!

  5. While not downplaying some of these issues, you really don’t know much about the history & existing knowledge in these areas. fMRI is, obviously, most sensitive to areas with larger vasculature & draining veins. Go back to the early to mid 1990’s articles and this is a good portion of what they focus on. You reference Chang & Glover, but Glover is a respected name in the field & he’s been studying cardiac & respiration noise & artifacts for a long time (for example search RETROICOR 1999) artifacts for a long time. Biswal is the first person to publish on resting connectivity (1996) It seems the core of Di… Biswal’s Ceb Cortex paper you site is that they’re using an easy-to-collect calibration method (just the standard deviation of a resting scan) rather than more complex methods like induced hypercapnia. The results are similar to the many hypercapnia studies. Check the references to the Di paper.

    As for your specific concerns regarding the insula & ACC, remember that some of these draining vein vascular artifacts are less on modern 3T scanners than on 1.5T scanners. If your sinister methodoligical confound was correct, we should have expected the late 90’s to be the explosion of research in these areas & around other heavily vascularized areas.

    There are many sloppy studies out there that might be more sensitive to these issues, but good task designs & cognitive subtraction goes a long way. There’s a lot of good work showing the resting connectivity has a real neural component, but the challenge of distinguishing neural connectivity differences from vascular or motion differences across populations is still a wide open and acknowledged research concern.

    There are reasons for the insula/ACC fad… and I agree it’s a bit of a fad, but all the research can’t be explained away as a simple vascular confound.

    • I’m familiar with the body of both Glover and Biswal’s work. I’m not sure what I said that caused you to think I was so ignorant of the history of vascular problems in MRI. I’m no methodologist, my primary work is in cognitive control, but my methods teacher happens to have a focus in respiration and noise filtering in MRI. I don’t really agree with your assertion that we should have seen this explosion earlier. If you look at the figure (which is obviously not of scientific quality, but for illustration purposes) you can see that the ACC spike started much earlier than the past two years.

      Further my exact criticism is that these issues, while probably not so much an issue for a well controlled within-subjects design, could severely impact both event-related and resting states using between subject design. I know in my own data, and my colleagues data, the anterior insula is one region that seems to pop in all of our data. Given that it’s clearly more susceptible to C02 artifacts, I don’t think we should hand wave away this issue.

      And you are correct; a lot of work has been done linking resting state to neural activity. I’m not one to claim resting state is all bullshit. But you have to admit that these problems are magnified immensely when one is dealing with groups of people sitting at rest for 5+ minute periods.

  6. Where in the brain is smell?
    No two papers on brain imaging and olfaction agree on anything. PET studies found greater right orbitofrontal cortex (OFC) activation with odour versus no odour (Small et al,1997). Some fMRI studies also report a right-sided asymmetry (O’Doherty et al, 2000; Sobel et al, 1998; Yousem et al, 1999). O’Doherty et al (2000), showed 6/8 had stronger right OFC activation than left. Using banana and vanillin as stimuli, found right OFC activation in 4/5 subjects and left OFC activation in 3/5. In contrast Zald and Pardo, (2000) found greater left activation with hydrogen sulphide. Left activation was also found by Royet et al (2000) using PET in response to pleasant and unpleasant olfactory stimuli. To add further confusion, Fulbright et al (1998), using fMRI, did not find any OFC activation when comparing pleasant and unpleasant odours – then, a subsequent study by Rolls et al (2003) found that pleasant odours activated the medio-rostral OFC whereas no activity was observed in the OFC for unpleasant odours. Both unpleasant and pleasant odours activated the anterior cingulate cortex and anterior insula (Rolls et al., 2003) but it is a medial activation and doesn’t appear to show marked lateralization with the exception that unpleasant odours activated a region of the right OFC. Levy et al (1997) using fMRI found that odour activated the OFC, cingulate gyrus, piriform and entorhinal cortices, hippocampus and amygdala. PET studies found vanillin activated the amygdala, piriform cortices bilaterally and parts of the anterior insula cortex – and caused frontal and right parietal deactivation in women (though not in men). Women exhibited less activation than men in Levy et al (1997) in contradiction to EEG studies in which women showed higher amplitude evoked responses and also to other fMRI studies (Yousem et al., 1999). PET studies showed no such gender differences (Bengtsson et al, 2001). The OFC has been implicated in hedonic differentiation with the activation in the right medial OFC being greater for pleasant than unpleasant odours but the left lateral OFC showed greater responsiveness to unpleasant than pleasant odours (Anderson et al, 2003). The bilateral middle orbitofrontal cortex (OFC), left lateral OFC, right insula, and bilateral anterior/middle cingulate gyri were most frequently activated by odor stimulation (Katata et al, 2009), but, in contrast to Anderson et al, the left middle OFC and right lateral OFC were significantly more often activated in the participants who perceived the odour stimulation as unpleasant, while the right anterior cingulate gyrus was more often activated in those who perceived the odour as pleasant.

    In the last few years I have looked at EEG and lateralisation of olfactory brain activation, following the lead of Davidson (1995). However, with the hegemony of brain imaging, and with the confusion outlined above, I haven’t had the confidence to publish opposing results.

    1. Anderson, A.K., Christoff, K., Stappen, I., Panitz, D., Ghahremani, D.G., Glover, G., Gabrieli, J.D.E., Sobel, N. 2003. Dissociated neural representations of intensity and valence in human olfaction. Nature Neuroscience 6 (2) , pp. 196-202

    2. Bengtsson, S., Berglund, H., Gulyas, B., Cohen, E., Savic, I. 2001. Brain activation during odor perception in males and females. NeuroReport 12 (9) , pp. 2027-2033

    3. Davidson, R.J., Sutton, S.K. 1995. Affective neuroscience: The emergence of a discipline. Current Opinion in Neurobiology 5 (2) , pp. 217-224

    4. Fulbright, R.K., Skudlarski, P., Lacadie, C.M., Warrenburg, S., Bowers, A.A., Gore, J.C., Wexler, B.E. 1998. Functional MR imaging of regional brain responses to pleasant and unpleasant odors. American Journal of Neuroradiology 19 (9) , pp. 1721-1726

    5. Katata, K., Sakai, N., Doi, K., Kawamitsu, H., Fujii, M., Sugimura, K., Nibu, K.-I. 2009. Functional MRI of regional brain responses to ‘pleasant’ and ‘unpleasant’ odors.Acta Oto-Laryngologica 129 (SUPPL. 562) , pp. 85-90

    6. Levy, L.M., Henkin, R.I., Hutter, A., Lin, C.S., Martins, D., Schellinger, D. 1997 .Functional MRI of human olfaction. Journal of Computer Assisted Tomography 21 (6) , pp. 849-856

    7. O’Doherty, J., Rolls, E.T., Francis, S., Bowtell, R., McGlone, F., Kobal, G., Renner, B., Ahne, G. 2000. Sensory-specific satiety-related olfactory activation of the human orbitofrontal cortex. NeuroReport 11 (4) , pp. 893-897

    8. Rolls, E.T., Kringelbach, M.L., De Araujo, I.E.T. 2003. Different representations of pleasant and unpleasant odours in the human brain. European Journal of Neuroscience 18 (3) , pp. 695-703

    9. Royet, J.-P., Zald, D., Versace, R., Costes, N., Lavenne, F., Koenig, O., Gervais, R. 2000. Emotional responses to pleasant and unpleasant olfactory, visual, and auditory stimuli: A positron emission tomography study. Journal of Neuroscience 20 (20) , pp. 7752-7759

    10. Small, D.M., Jones-Gotman, M., Zatorre, R.J., Petrides, M., Evans, A.C. 1997. Flavor processing: More than the sum of its parts. NeuroReport 8 (18) , pp. 3913-3917

    11. Sobel, N., Prabhakaran, V., Desmond, J.E., Glover, G.H., Goode, R.L., Sullivan, E.V., Gabriell, J.D.E. 1998. Sniffing and smelling: Separate subsystems in the human olfactory cortex. Nature 392 (6673) , pp. 282-286

    12. Yousem, D.M., Maldjian, J.A., Siddiqi, F., Hummel, T., Alsop, D.C., Geckle, R.J., Bilker, W.B., Doty, R.L. 1999. Gender effects on odor-stimulated functional magnetic resonance imaging. Brain Research 818 (2) , pp. 480-487

    13. Zald, D.H., Pardo, J.V. 2000. Functional neuroimaging of the olfactory system in humans. International Journal of Psychophysiology 36 (2), pp. 165-181

  7. As many of you had questions or comments regarding the best way to deal with respiratory related issues, I spoke briefly with Torben at yesterday’s lab meeting. Removal of respiratory noise is fairly simple, but the real problem is with end-tidal C02. According to Torben, most noise experts agree that regression techniques only partially remove the artifact, and that an unknown amount is left behind even following signal regression. This may be due to slow vascular saturation effects that build up and remain irrespective of shear breath frequency. A very tricky problem indeed.

    • Hi Micah

      What do you mean when you state that “removal of respiratory noise is fairly simple”? Do you mean that the mathematical operation is simple or well understood? Do you mean that it is effective? Although it is nice when a mathematical operation is simple and/or well understood it doesn’t amount to a hill of beans if it isn’t effective. So where is the evidence that the filtering of respiratory noise by present means is effective?

      And by the way, what is the accepted means of removing the respiratory noise? Is it through band-pass filtering of Fourier frequency components? Is it through wavelet filtering of scale and temporal focality? Is it through statistical methods which attempt to account for modeled confounds and associated measurements?

      • Hi Daniel,

        Thanks for your comment. After further discussion with Torben, I should probably remove my comment that noise filtration is simple, as it’s anything but. While the common methods for dealing with it are fairly straight forward- for example adding a column to the design matrix to model respiration-related effects- such approaches cannot account for residual artifacts relating to spin-history effects and end-tidal C02. While regression of respiration-related movement is fairly simple, it appears to often underperform and can often strip important signal from your data. Band-pass filtering is one common standby which removes high and low frequency noise. However it’s not clear that saturation effects from end-tidal C02 are covered by this method; Torben says that even when attempting to remove the noise via the GLM, a good deal of noise signal is left behind. The truth is, our best methods seem to make little difference at best, and at worst will totally strip all interesting signal from your data. So the problem is far from solved and various techniques all come with harsh trade-offs.


  8. Thanks for the great write up. In light of this, do you need to revise the conclusions in your JN 2012 paper which makes claims about the MPFC and AIC?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.