A Scuffle In The Coma Ward

A couple of months ago, the BBC TV show Panorama covered the work of a team of neurologists (led by Prof. Adrian Owen) who are pioneering the use of fMRI scanning to measure brain activity in coma patients.

The startling claim is that some people who have been considered entirely unconscious for years, are actually able to understand speech and respond to requests - not by body movements, but purely on the level of brain activation.

However, not everyone was impressed. A group of doctors swiftly wrote a critical response, published in the British Medical Journal as fMRI for vegetative and minimally conscious states: A more balanced perspective

The Panorama programme... failed to distinguish clearly between vegetative vs. minimally conscious states, and gave the impression that 20% of patients in a vegetative state show cognitive responses on fMRI.

There are important differences between the two states. Patients in a vegetative state have no discernible awareness of self and no cognitive interaction with their environment. Patients in a minimally conscious state show evidence of interaction through behaviours...

The programme presented two patients said to be in a “vegetative state” who showed evidence of cognitive interaction on assessment using fMRI but the clinical methods used for the original diagnosis were not stated. In both cases, family members clearly reported that the patient made positive but inconsistent behavioural responses to questions... one of these patients was filmed responding to a question from his mother by raising his thumb and the other seemed to turn his head purposefully.

So Panorama stands accused of passing off patients who were really minimally conscious, as being in a vegetative state. To see signs of understanding on brain scans from the latter would be truly amazing because it would be the first evidence that they weren't, well, vegetative.

However if they were 'merely' minimally conscious patients, it's not as interesting, because we already knew they were capable of making responses.

Now the Panorama team - and Professor Owen - have replied in a BMJ piece of their own. Given that they're charged with  misleading journalism and sloppy medicine, they're understandably a bit snarky:

Just by viewing this one hour documentary the authors felt able to discern that both the patients “said to be in a vegetative state” are “probably” minimally conscious... One of these patients, Scott, has had the same neurologist for more than a decade. Professor Young, who appeared in the film, made it clear that Scott had appeared vegetative in every assessment...

The fact that these authors took Scott’s fleeting movement, shown in the programme, to indicate a purposeful (“minimally conscious”) response shows why it is so important that the diagnosis is made in person, by an experienced neurologist, using internationally agreed criteria.

In other words, they were vegetative, and the critics who said otherwise, on the basis of some TV footage, were being silly.

In other words...it's on.

Turner-Stokes L, Kitzinger J, Gill-Thwaites H, Playford ED, Wade D, Allanson J, Pickard J, & Royal College of Physicians' Prolonged Disorders of Consciousness Guidelines Development Group (2012). fMRI for vegetative and minimally conscious states. BMJ (Clinical research ed.), 345 PMID: 23190911

Walsh F, Simmonds F, Young GB, & Owen AM (2013). Panorama responds to editorial on fMRI for vegetative and minimally conscious states. BMJ (Clinical research ed.), 346 PMID: 23298817

Drunk Rats Could Overturn Neurological Orthodoxy

A form of brain abnormality long regarded as permanent is, in fact, sometimes reversible, according to an unassuming little paper with big implications.

Here's the key data: some rats were given a lot of alcohol for four days (the "binge"), and then allowed to sober up for a week. Before, during and after their rodent Spring Break, they had brain scans. And these revealed something remarkable - the size of the rats' lateral ventricles increased during the binge, but later returned to normal.

Control rats, given lots of sugar instead of alcohol, did not show these changes.

This is really pretty surprising. The ventricles are simply fluid-filled holes in the brain. Increased ventricular size is generally regarded as a sign that the brain is shrinking - less brain, bigger holes - and if the brain is shrinking that must be because cells are dying or at least getting smaller. So bigger ventricles is bad.

Or so we thought... but this study shows that it might not always be true: alcohol reversibly increases ventricular volume over a timescale of days. It does so, the authors say, essentially by drying brain tissue out; like most things, if you dry the brain out, it gets smaller (and the ventricles get bigger) but when the water comes back to the tissues, it expands again.

As you can see here in Figure 2...

Maybe. I admit that just eyeballing this, it looks more like the ventricles are getting brighter, rather than bigger, but I'm not familiar with the details of water scanning. Maybe some readers will know more about it.

If it's true, this is big - maybe it's not just high doses of alcohol that does this. Maybe other drugs or factors can shrink or expand, the ventricles, or even other areas, purely by acting on tissue water regulation, rather than by anything more 'interesting'.

Take the various claims that some psychiatric drugs boost brain volume while others decrease it, just for starters...could they be headed for a watery grave?

Of course, this is in mice - and it might not translate to humans... we need to find out, and I for one am keen to apply for a grant. Here's my draft:

Participants: 8 healthy-livered neuroscientists.
Materials: 1 MRI scanner, 1 crate Jack Daniels.
Methods: Subjects will confer to pick a Designated Operator, who will remain sober. If no volunteers for this role are forthcoming, selection will be randomized by Bottle Spinning. All other participants will consume Jack Daniels ad libitum, and take turns being scanned. Once all Jack Daniels is depleted, participants will continue to be scanned until fully sobered up (defined as when they can successfully spell "amygdalohippocampal").
Instructions to Participants: i) what happens in the magnet, stays in the magnet. ii) If you 'dirty' the scanner, you clean it up. iii) Bottle caps are not MRI safe!

Er... seriously though, someone should check.

Zahr NM, Mayer D, Rohlfing T, Orduna J, Luong R, Sullivan EV, and Pfefferbaum A (2013). A mechanism of rapidly reversible cerebral ventricular enlargement independent of tissue atrophy. Neuropsychopharmacology  PMID: 23306181

Flawed Statistics Make Almost Everyone's Brain "Abnormal"

A popular method for detecting abnormalities in the shape and size of individual brains is seriously flawed, and is almost guaranteed to find 'differences' even in normal people.

So say Italian neuroscientists Scarpazza and colleagues in an important new report: Very high false positive rates in single case Voxel Based Morphometry.

Voxel Based Morphometry (VBM) is a way of analyzing brain scans to detect structural differences. It's most commonly used to compare groups of brains to find average differences, but some neuroscientists have started using VBM to check for abnormalities in a single brain. Scarpazza et al list 34 pieces of research about that, including 13 since 2010.

So it would suck if there were a problem with individual VBM... but there is. This pic tells the tale:

The authors took 200 normal brains and compared each one of them in turn to a control group of 16 normal brains. Because all of them were healthy, the comparisons ought to show no significant differences.

The technique was set up so that, in theory, only 5% of the brains should have been wrongly labelled as containing an abnormality. But in fact, a full 93.5% of the normal brains gave at least one false positive.

So 5% is more like the rate of not being wrong. Oops.

The image shows that in some brain areas, almost 25% of the normal brains were branded as 'abnormal' just in that region alone - the hotter the colour, the higher the proportion of false 'hits'. The top row is for false reports of brain volume increases, while the bottom row is decreases; false 'increases' were more common.

So what's going wrong? It's not entirely clear and several factors are probably at play, but the authors say that the main issue is that VBM makes the assumption of statistical normality which doesn't in fact hold.

Either way, it's a serious problem, and Scarpazza et al point to one especially worrying implication: some people have proposed using single-subject VBM in a legal context, to reinforce insanity pleas by showing subtle 'brain abnormalities' not obvious to the naked eye. Yet if this paper's right, such evidence could be entirely meaningless, almost guaranteed to give a positive result.

P.S. Last time I posted about this kind of analysis flaw, the internet went crazy because they didn't understand it. So just to be clear, this is not a problem for clinical scans - the kind you'd get to check whether you have a brain tumour.

Scarpazza, C., Sartori, G., De Simone, M., and Mechelli, A. (2013). When the single matters more than the group: Very high false positive rates in single case Voxel Based Morphometry NeuroImage DOI: 10.1016/j.neuroimage.2012.12.045

How Intelligent is IQ?

"If your IQ is somewhere around 60 then you are probably a carrot'', according to a British spokesman for high-IQ club Mensa.

IQ's in the news at the moment thanks to a paper called Fractionating Human Intelligence from Canadian psychologists Adam Hampshire and colleagues. Some say it 'debunks the IQ myth' - but does it?

The study started out with a huge online IQ test...

Behavioral data were collected via the Internet between September and December 2010. The experiment URL was originally advertised in a New Scientist feature, on the Discovery Channel web site, in the Daily Telegraph, and on social networking web sites including Facebook and Twitter.

The test involved 12 different cognitive tasks, based on the usual IQ test kind of things, and they got a huge 45,000 usable responses.

However, the main part of the study used functional MRI (fMRI) to measure brain activity caused by each of the 12 tasks. There were only 16 volunteers in the brain scan study, which is pretty small.

The key finding was that although each of the 12 tasks made a different pattern of brain regions light up, there were two main components of this: one lit up mostly in response to tasks requiring short-term memory, and the other was associated with reasoning and logic: (EDIT: Picture corrected, oops.)

They did various other analyses that confirmed this, and they also found evidence for a third network responsible for language (verbal) skill.

Finally, the killer conclusion was that there was no reason to introduce the imfamous  'g factor' - a number representing general intelligence affecting performance on all tasks. Although there was a 'g factor' statistically, it was explained by the fact that tasks required both the memory and the logic networks (although to different degrees).

g is the most controversial aspect of IQ testing, because if it exists, that means that some people are just smarter than others across the board - not just better at a particular kind of thing. So has this study killed g?

Well, not by itself. There's a huge literature on IQ and g, going back almost 100 years. This stuff is not based on brain imaging, but just on IQ test scores, and it's a complex topic. I don't think one brain study with 16 people can really overturn that, although it does lend weight to the anti-g camp who have been arguing against g for decades.

There's a sense, though, in which it doesn't matter. If all tasks require both memory and reasoning (and all did in this study), then the sum of someone's memory and reasoning ability is in effect a g score, because it will affect performance in all tasks.

If so, it's academic whether this g score is 'really' monolithic or not. Imagine that in order to be good at basketball, you need to be both tall, and agile. In that case you could measure someone's basketball aptitude, even though it's not really one single 'thing'...

Hampshire, A., Highfield, R., Parkin, B., and Owen, A. (2012). Fractionating Human Intelligence Neuron, 76 (6), 1225-1237 DOI: 10.1016/j.neuron.2012.06.022

The Perils of Sharing Brain Scans

A fascinating paper by neuroscientists Van Horn and Gazzaniga chronicles their pioneering, but not entirely successful, attempt to get researchers sharing their brain scans: Why share data? Lessons learned from the fMRIDC.

It all started in 1999 when, along with some colleagues, they decided that the time was right for data sharing in neuroimaging. They got some public funding, and tried to get various major neuroscience journals to require that anyone publishing an fMRI study should make their data available to the fMRI Data Consortium (fMRIDC).

By making it mandatory, they'd ensure that there was no selection bias. Requirements to post raw data were already common in other fields of science like genetics and crystallography. So, they thought, why can't it happen here?

However, it didn't go down very well:

Upon becoming aware of our efforts and goals, fMRI researchers angered by journal requirements to provide copies of the fMRI data from their published articles began a letter writing campaign seeking to muster opposition  an effort which was featured in the news and editorial sections of several influential journals.

Editorials and commentaries over fMRI data sharing were aired in the pages of Science, Nature Neuroscience etc. expressing concern over the data sharing requirement, over what possession of the data implied, human subject concerns, and, if databasing was to be conducted at all, how it should be conducted “properly”.

This was all before my time, sadly. It sounds like a grand old academic street-fight. No doubt those on the other side remember it differently from how it's presented here, though.

The reactions of our colleagues caught us somewhat off guard. We were honestly surprised by the  negative and hostile response when we had believed that creation of a data archive would be of benefit to the neuroimaging community. Perhaps, they had a point.

Maybe the field wasn't ready for fMRI data sharing? Perhaps, it was too early to start such a project? We struggled with how best to move forward or whether to move forward at all.

Anyway, they decided they would continue, on a more modest scale. fMRIDC ended up with data from about 100 fMRI studies by the time the funders pulled the plug in 2006, only a fraction (and perhaps an unrepresentative one) of the papers published, but still, it's something.

As I said, I missed out on this debate, but if I had been in the field 10 years ago, I suspect I'd have been on the extreme wing of the pro-sharing faction, the Montagnard to the founders' Girondism. My view is that no researcher owns their data. The only person who owns the results of a brain scan is the person whose brain it is.

If I could wave a magic wand, I'd put a chip in every MRI scanner that automatically uploaded all scans to a public database as soon as they appeared (with the subject's consent, and/or with personal information about the subject stripped out). I'd fix all the software such that every time someone ran an analysis, it was publicly logged. Total transparency is best for science, I believe, and it would also make scientists lives  easier, once they got over the initial shock.

Sadly that's not possible... yet... but data sharing is a noble cause and, as Van Horn and Gazzaniga point out, even if fMRIDC is dead, the idea lives on with many new initiatives emerging, hydra-like, in its place.

Van Horn JD, and Gazzaniga MS (2012). Why share data? Lessons learned from the fMRIDC. NeuroImage PMID: 23160115

Autism Brain Scans Flawed? You Read It Here First

According to a piece in Nature today, a major line of research about autism might be seriously flawed:

One of the most popular and widely accepted theories on the cause of autism spectrum disorders attributes the condition to disrupted connectivity between different regions of the brain.

This 'connectivity hypothesis' claims that the social and cognitive abnormalities in people with autism can be explained by a dearth of connections between distant regions of the brain. Some flavours of this theory also predict more connections between nearby brain regions.

Recent studies, however, have found that when a person moves their head while undergoing functional magnetic resonance imaging (fMRI) - a method that maps how different neuroanatomical structures of the brain interact in real time, its functional connectivity - it looks like the neural activity observed in autism. That's a sobering discovery...

So the characteristic pattern of "abnormal connectivity" in autism might not be real: it might just reflect the fact that people with autism move around more during the scan.

A sobering idea, indeed... but not an entirely new one. I suggested it over a year ago:

Head motion affects estimates of functional connectivity. The more motion, the weaker the measured connectivity in long-range networks, while shorter range connections were stronger... Disconcertingly, this is exactly what's been proposed to happen in autism (although in fairness, not all the evidence for this comes from fMRI). This clearly doesn't prove that the autism studies are all dodgy, but it's an issue. People with autism, and people with almost any mental or physical disorder, on average tend to move more than healthy controls.

Ben Deen, and Kevin Pelphrey (2012). Perspective: Brain scans need a rethink Nature

Is fMRI About To Get Fifty Times Faster?

According to a paper just published, a new technique of functional MRI scanning (fMRI) could soon allow neuroscientists to measure brain activity far faster: Generalized iNverse imaging (GIN): Ultrafast fMRI with physiological noise correction

Authors Boyacioglu and Barth claim remarkable things for the technique:

We find that the spatial localization of activation for GIN is comparable to an EPI protocol and that maximum z-scores increase significantly... with a high temporal resolution of 50 milliseconds.

EPI, the current standard fMRI sequence, would have a temporal resolution of 2000 or 3000 milliseconds, so it's about 50 times faster.

Other super-fast fMRI methods already exist (e.g. this one I blogged about), but they've generally achieved speed only at a cost: they've had to either sacrifice spatial resolution to achieve that, or limited themselves to scanning only a small fraction of the brain, or have been more subject to random noise and hence less sensitive.

GIN, however, is said to cover the whole brain, with decent spatial resolution and signal-to-noise ratio. The data can be analyzed in exactly the same way as any other kind. So that's up to fifty times faster with no real drawbacks.

That would be truly revolutionary - as the major limitation of fMRI at the moment is that it's much slower than other methods of recording brain activity.

Check it out: this shows brain activation in response to simple visual stimuli, imaged with bog-standard EPI and GIN:

So this is a big deal... if it does work, I'm sure neuroscientists the world over will be lining up to buy Boyacioglu and Barth a GIN and tonic.

How does it work, and is it all it's cracked up to be? Well, I can't really say: the math is beyond me.

In essence, rather than scanning the brain in 3D, slice by slice (like this), GIN only scans one 2D slice, but then manages to reconstruct the rest of the brain in 3D from just that slice, using dark, forbidden magicks... I mean mathematics. The principle is called parallel imaging and it's been around for several years, but with image quality limitations that GIN claims to have overcome.

Perhaps my more technically-inclined readers will have more insightful comments.

Boyacioglu R, and Barth M (2012). Generalized iNverse imaging (GIN): Ultrafast fMRI with physiological noise correction. Magnetic Resonance in Medicine PMID: 23097342

More on False Positive Neuroimaging

Back in June, I warned that the ever-increasing number of clever methods for analyzing brain imaging data could be a double-edged sword:

Recently, psychologists Joseph Simmons, Leif Nelson and Uri Simonsohn made waves when they published a provocative article called False-Positive Psychology - Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant.

It explained how there are so many possible ways to gather and analyze the results of a  simple psychology experiment that, even if there's nothing interesting really happening, it'll be possible to find some "significant" positive results purely by chance...

The problem's not just seen in psychology however, and I'm concerned that it's especially dangerous in modern neuroimaging research.

In a comment on that post, The Neurocritic pointed out that Michigan PhD student Joshua Carp had put forward the same argument in a conference presentation, several months previously.

Now Carp's published a paper on the topic: On the plurality of (methodological) worlds: estimating the analytic flexibility of fMRI experiments. It's free to access, so check it out.

Whereas I just talked the talk by listing lots of possible ways in which you could analyze a given set of data, Carp walked the walk, and actually did loads of analyses. He took a single dataset, the results of a simple experiment and looked at it in almost 7000 different ways. Each set of results was then thresholded to correct for multiple comparisons in 5 ways, for a grand total of 35,000 outputs.

The variants he considered ranged from how much smoothing to apply, to how to correct for head motion, and many more.

What happened? In a nutshell, the different options made a difference - and the variability was the largest in parts of the brain that were most activated (the "blobs" that lit up). In other words, analytic flexibility makes the most difference in the most interesting places. See the picture at the top.

The location of the maximum peak activation also varied. This is not unexpected, and not, in itself, that worrying - the great majority of the peaks clustered in a few small areas. However, it underlines that different options really can make a difference.

Carp concludes:

Nearly every voxel in the brain showed significant activation under at least one analysis pipeline. In other words, a sufficiently persistent researcher determined to find significant activation in virtually any brain region is quite likely to succeed...

If investigators apply several analysis pipelines to an experiment, and only report the analyses that support their hypotheses, then the prevalence of false positive results in the literature may far exceed the nominal rate. However, analytic flexibility only translates into elevated false positive rates when combined with selective analysis reporting. If researchers reported the results of all analysis pipelines used in their studies, then it would not be problematic.

To the author’s knowledge, there is no evidence that fMRI researchers actually engage in selective analysis reporting. But researchers in other fields do appear to pursue this strategy.

In my experience, fMRI researchers are actually fairly conservative in terms of using different analyses, and certainly I doubt anyone has ever run thousands of them just to get the result they want and I'd estimate that most published findings are not the result of more than a handful of 'attempts' at most.

However it's a serious concern that it could happen, and importantly it's getting ever-easier to do this, with the continuing increase in computer power making running an analysis quicker and cheaper than ever. As to what to do about it, Carp makes several suggestions, and here's one I made earlier...

Joshua Carp (2012). On the plurality of (methodological) worlds: estimating the analytic flexibility of fMRI experiments Front. Neurosci. DOI: 10.3389/fnins.2012.00149

A New Theory of Psychosis?

A team of British neuroscientists led by the (in)famous David Nutt says that magic mushrooms offer a new theory of psychosis: Functional Connectivity Measures After Psilocybin Inform a Novel Hypothesis of Early Psychosis

It's a reanalysis of a study from earlier this year, which got quite a lot of attention, in which 15 volunteers were injected with psilocybin - the major active hallucinogenic ingredient in 'magic mushrooms' - during an fMRI scan.

In a nutshell, the rather interesting proposal in the new paper is that psilocybin may cause mind-altering effects by blurring the difference between the brain networks responsible for 'internal' and 'external' thought.

Activity in the internal "default mode network" (DMN) is generally anti-correlated with the "task-positive network" (TPN) - when one is higher, the other's lower. The DMN is active when you're not doing much - hence 'default' while the TPN comes online when you're engaged in a particular mental activity.

Nutt's team say, however, that their functional connectivity fMRI data show that after psilocybin, activity in these two networks becomes positively correlated - an unusual pattern. They write:

Increased DMN-TPN coupling has been found in people at high risk of psychosis and an inability to distinguish between one’s internal world and the external environment, sometimes referred to as “disturbed ego boundaries,” is a hallmark of early psychoses and the psychedelic state.

One of our volunteers reported the following after psilocybin: “It was quite difficult at times to know where I ended and where I melted into everything around me.”

To be honest I'd need to see a replication before I put too much faith in this, because this kind of post-hoc reanalysis of fMRI data is very flexible and therefore prone to false positives, but it's an interesting idea, and at least it provides a clear theory for further research.

Carhart-Harris RL, Leech R, Erritzoe D, Williams TM, Stone JM, Evans J, Sharp DJ, Feilding A, Wise RG, and Nutt DJ (2012). Functional Connectivity Measures After Psilocybin Inform a Novel Hypothesis of Early Psychosis. Schizophrenia bulletin PMID: 23044373

Control A Robot With Your Brain?

A paper just out makes the dramatic claim that you can control a robot using thought alone, Avatar style, thanks to a 'mind reading' MRI scanner. But does it really work?

Dutch neuroscientists Patrik Andersson and colleagues bought a robot - an off-the-shelf toy called the 'Spykee' -  which is equipped with Wifi and  a video camera. The controlling human lay in the scanner and real-time fMRI was used to record brain activity. The video feed from the robot was showed on a screen in the scanner, completing the human-robot loop.

Participants controlled the robot with their brain. Specifically, they had to focus their attention on one of three arrows - forward, left, and right - shown on the screen.

During an initial training phase they focussed on each arrow in turn, to provide examples of the resulting brain activity: these were then fed into a machine learning algorithm that learned to recognize the pattern of BOLD activation for each command. Then in the second phase, they could control the robot just by thinking about the correct arrow - the scanner 'decoded' their brain activity and sent the appropriate commands to the bot over Wifi.

None of the elements of this process are new - real time fMRI has been around for a few years, so has machine learning to decode brain activation - but it's the first time they've been put together in this way.

And it's pretty awesome. The participants were able to guide their 'avatar' around a room to visit a number of target locations. They weren't perfectly accurate, and it took 10 or 15 minutes to navigate a few meters of ground... but it worked.

However... were they really using their minds, or just their eyes?

This is my main concern about this paper: participants were told to keep their eyes focussed on the middle of the screen and just mentally focus on the arrows to give commands. If they did indeed keep their eyes entirely stationary, then the patterns of brain activation would indeed represent pure 'thoughts'.

But if they were moving their eyes slightly (even unconsciously), the interpretation would be rather different. Moving their eyes would change the pattern of light hitting their retina, and this would be expected to change brain activation in the visual system of the brain.

So, maybe the fancy fMRI decoding system wasn't reading their mind, it was just acting as an elaborate means of tracking eye movements - which would be much less interesting. If you want to control a robot with your eyes, there are cheaper ways.

Andersson et al acknowledge this issue, and they claim, for various reasons, that this probably wasn't what happened here - but they didn't measure eye movements directly, so it does remain a worry. Eye tracking devices suitable for fMRI are widely available but this study used an ultra-powerful 7 Tesla scanner which, the authors say, made it impossible. So there's more work to be done here.

Andersson P, Pluim JP, Viergever MA, and Ramsey NF (2012). Navigation of a Telepresence Robot via Covert Visuospatial Attention and Real-Time fMRI. Brain topography PMID: 22965825

When Data Filtering Introduces Bias (fMRI Edition)

A couple of months ago I blogged about a paper showing that 'filtering' of EEG data can create spurious effects.

Now, we read about another form of bias that filters can introduce, this time for fMRI: Filtering induces correlation in fMRI resting state data.

Australian neuroscientists Catherine Davey and colleagues consider temporal filtering of fMRI data in studies looking at correlation (brain functional connectivity).

Because both very high frequency and very slow changes in the fMRI signal are probably caused by artefacts, rather than interesting brain signals, it's common to use a filter to try and extract the medium-frequency changes that are of most interest (e.g. approximately 0.01 to 0.1 Hz).

However, while this filtering is very useful, Davey et al show that it can - ironically - create artefacts of its own: here's the data from one volunteer scanned during a simple task and then analyzed in 4 different ways:

Without filtering (A) there's a huge amount of 'connectivity' - too much to be realistic. This is why filtering is important.

But filtering, without correcting for the effects of the filter, actually makes things worse (B). It solves one problem but at the cost of creating another. The problem is those pesky autocorrelations. The authors say, however, that they've calculated a way to correct for filter-induced correlations (D) and that this gives more realistic results. They recommend that this should be used in future connectivity studies, but don't go into much detail regarding the question of what this means for the existing literature.

Perhaps data 'filtering' is a misleading term. It implies that all you're doing is removing the unwanted noise, leaving pristine, crystal clear data, a bit like a water filter. Mmm. What could go wrong? In fact mathematical 'filters' can put stuff into the data as well as take it out, so should we stop using that word and just call them what they are: modifications?

Davey CE, Grayden DB, Egan GF, and Johnston LA (2012). Filtering induces correlation in fMRI resting state data. NeuroImage PMID: 22939874

This Is Your Brain On Management

Have you ever wondered whether how the brains of managers work? New research from a group of German neuroscientists and management experts reveals all: Dissociated Neural Processing for Decisions in Managers and Non-Managers

The results were rather remarkable:

Using fMRI, the researchers found that managers' brains were less active in a number of areas, compared to the brains of non-managers, when doing the same task. By contrast, managerial brains were more active than the others only in one small area (caudate nucleus). See above.

So overall, managers had less brain activation during the task. Does that mean they have defective brains? Could this be a neurobiological explanation for the behaviour of Pointy Haired Boss and David Brent?

Not at all, say the authors. The lower activation in the brains of managers means that they were more efficient:

the managers might have found a more efficient way of sorting the presented words. This might have enabled them to faster decide for their preferred category...  Managers as expert decision-makers would seek to find a rule or heuristic on which they could base their decisions. According to previous studies, this phase of rule identification would involve the caudate nucleus

While non-managers wasted brainpower on thinking through the task with several areas of their cerebral cortex, the managers (so to speak) downsized their neurological expenditure by outsourcing the work to their caudate nucleus, an area responsible for applying a simple but effective rule.

One of the problems with these kinds of group-comparison fMRI studies is that under-activation can equally well be glossed as "deficient" or "efficient". Curiously, it usually ends up being whichever fits with the author's narrative.

That's assuming you agree that the task was about "decision making". It consisted of seeing a long series of pairs of words, one "individualistic" such as 'power' and one "collectivistic" such as 'harmony'. Participants just had to pick which word they liked best. There were no right or wrong answers. I'm not sure what kind of manager would have to do anything like that in real life. Maybe a manager of a fridge magnet poetry manufacturer?

That's also assuming the results are solid. The authors provide few details on the fMRI methods (the main results are said to be "cluster-level corrected at p less than 0.0013", which is an unusual threshold to use and an extremely strict one (0.05 cluster-level is more common; this is about 40 times stricter).

Still. If you do buy these results, the message is: management is literally about using as little of your brain as possible...

Caspers S, Heim S, Lucas MG, Stephan E, Fischer L, Amunts K, and Zilles K (2012). Dissociated neural processing for decisions in managers and non-managers. PloS one, 7 (8) PMID: 22927984

Neuroscience: Solving The Hard-On Problem

Take a good look at this, fellas:

It might not look like much, but this is science's very first glimpse of something rather close to the hearts of most men.

These images show nerve activation in the spinal cord during sexual arousal. Using functional magnetic resonance imaging (fMRI), best known as a way of recording brain activity, applied to the spine, Canadian researchers Kozyrev et al were able to record the changes associated with, well, stimulation. Here's the paper: Neural correlates of sexual arousal in the spinal cords of able-bodied men.

The methods sound like a real riot. Ten guys were scanned while being exposed to two different kinds of stimulation. First, 'audiovisual'

Films, depicting heterosexual couples engaged in sex play, vaginal sexual intercourse, and oral sex, were presented in two 5-min blocks, separated by a baseline condition (blank screen) for 3 min... The selected films were chosen by the investigators and were all very highly rated (on a scale of 1 to 10) by independent viewers. Only films that received a minimum score of 8 out of 10 were shown to participants.

They don't say who these independent grumble-rating viewers were, but I'm, er, told that some porn sites these days have user ratings next to each clip. So it's possible that they used those and that the raters were, quite literally, just some bunch of wankers.

As for the manual stimulation...

All participants manually performed stimulation of the penis. Participants received written instructions superimposed onto a rear-projection screen, indicating whether to rest or to perform penile self-stimulation... there were two equal blocks of 1.5 min of stimulation that were separated by a 1.5-min block of rest... a shorter duration was selected for this part of the study in order to avoid overstimulation of the penis and/or ejaculation.

Good thinking. That could have been messy.

The results of the study largely confirmed previous work, but it's an important step forward because until now, no-one had measured arousal-related spinal activation directly in healthy humans; the evidence came from recordings in animals and spinal injury patients.

Clearly there's more work to be done here. The images are rather crude. That grey oval is meant to look like this:

Still, it's early days. Spinal cord fMRI is not a new technique, having first been performed in 1999, but it's not received nearly as much research interest as the brain variety.

Kozyrev N, Figley CR, Alexander MS, Richards JS, Bosma RL, & Stroman PW (2012). Neural correlates of sexual arousal in the spinal cords of able-bodied men: a spinal FMRI investigation. Journal of sex & marital therapy, 38 (5), 418-35 PMID: 22900624

Brains In Motion Are Bad For Neuroscience

A new paper in Human Brain Mapping reports on: Functional magnetic resonance imaging movers and shakers: Does subject-movement cause sampling bias?

Head movement is a well known problem that can badly impact the quality of neuroimaging data, introducing spurious signals and obscuring real ones. It's an issue for all brain scanning research but according to Wylie and colleagues, authors of this paper, it's especially serious for studies comparing disease patients to healthy controls.

The authors got 34 people with multiple sclerosis (MS) to perform some simple cognitive tasks during fMRI scanning. They found that the harder the task was, the more the patients moved around during the scan; and in those with more severe MS, the correlation between difficulty and motion was even stronger. In healthy people, harder tasks only caused slightly more motion.

And the more people moved, the less brain activity was recorded, probably because movement degraded the data quality.

fMRI data associated with severe head movement is frequently discarded. In discarding these data, it is often assumed that head-movement is a source of random error, and that data can be discarded from subjects with severe movement without biasing the sample. We tested this assumption by examining whether head movement was related to task difficulty and cognitive status among persons with multiple sclerosis (MS).

[In the MS patients] there was a linear increase in movement as task difficulty increased that was larger among subjects with lower cognitive ability. Analyses of the signal-to-noise ratio (SNR) confirmed that increases in movement degraded data quality. Similar, though far smaller, effects were found in healthy control subjects. Therefore, discarding data with severe movement artifact may bias multiple sclerosis samples such that only those with less-severe cognitive impairment are included in the analyses. However, even if such data are not discarded outright, subjects who move more will contribute less to the group-level results because of degraded SNR.

Oh dear. fMRI researchers use two main ways to deal with motion - correction, and rejection. Either you try to take account of movement and analyze the data, or you just throw out the results from people who move a lot. Most people use a combined approach, chucking out the really heavy movers and then using correction on the rest.

This paper however shows that both techniques have problems. Despite motion correction, data from heavy movers has a lower signal to noise ratio, so if you include them, they will "dilute" your sample. However, if you chuck them, that'll also introduce bias, because heavy movement is not random - people with more severe MS move more, so by excluding heavy movers, you'll be excluding severe cases.

Informally, every neuroimaging researcher knows that some people move more than others. Learning to spot likely "movers" and avoid wasting money on scanning them is a fine art. In my experience just about every "patient" population move, on average, more than healthy controls, and children and the elderly move more than young adults.

I'm not sure there's an ideal solution but perhaps the best approach is to run all analyses (at least) twice, once including everyone, regardless of movement, and then again, with strict movement exclusion criteria. Results consistent across both analyses are probably solid.

Wylie GR, Genova H, Deluca J, Chiaravalloti N, and Sumowski JF (2012). Functional magnetic resonance imaging movers and shakers: Does subject-movement cause sampling bias? Human brain mapping PMID: 22847906

Brain Scanning... Or Vein Scanning?

Many fMRI studies of brain activity could be biased by the effect of large blood vessels, according to an interesting new report: Origins of intersubject variability of BOLD and arterial spin labeling fMRI.

fMRI measures BOLD, the Blood Oxygenation Level Dependent response. As the name says, BOLD is when a bit of the brain becomes more active, it uses more oxygen, and the oxygenation level of the blood in the area drops - although it then increases to compensate, and it's the increase that most fMRI picks up.

There's a catch though: blood flows. Specifically, it flows from arteries, into tissues - the brain, in this case - and then into veins. Blood leaving the brain tends to end up in the larger veins and, being large, these exert a large effect on BOLD - even though they're some distance from the true site of neural activation.

So, the worry is that BOLD blobs may be shifted towards the nearest large vein, reducing the accuracy of fMRI. It's a well-recognized issue, but it's not clear just how serious it is... or what we can do about it.

Enter Canadian researchers Ismael Gaxiola-Valdez and Bradley Goodyear. By comparing BOLD to the alternative method of ASL - They show that the large vessel effect does happen to BOLD signals, shifting the activation blobs nearer to the surface of the brain (where veins are.) However, the effect is more pronounced when the blobs represent absolute BOLD changes. When blobs represent the z score - which they usually do - the difference is smaller.

That's somewhat reassuring... but here's the bad news. Some peoples' brains show larger BOLD changes than others. Gaxiola-Valdez and Goodyear show that these individual differences in BOLD magnitude are probably driven by large vessel effects because as the pic at the top shows, BOLD was most variable (between people) at the surface, i.e. near the veins, while ASL, less subject to the vein bias, was most variable deep in the brain, as you'd expect.

Worse, the size of people's BOLD and ASL signal changes (to the same task) were not correlated (or only weakly) when the BOLD study used short task blocks. This calls into question short-block or event-related fMRI experiments that measure differences between people or between groups: the differences might be in the veins, not the brain. With longer task blocks, however, there was a good correlation with the ASL.

Overall, it's an important paper and a reminder that, although neuroscientists sometimes treat the brain as separate from the rest of the body, it's not.

Gaxiola-Valdez I, and Goodyear BG (2012). Origins of intersubject variability of blood oxygenation level dependent and arterial spin labeling fMRI: implications for quantification of brain activity. Magnetic resonance imaging PMID: 22795932

BOLD Blobs Brighten Baby Brains

Babies born prematurely show the same kind of brain activation seen in adults: but it's a lot slower. That's according to an interesting study using fMRI scanning.

The authors, Tomoki Arichi and colleagues of London, measured brain activation in response to mild sensory stimulation (touching the right hand) in three groups: adults, "preterm" infants who were just 38 weeks old since conception, and "term" infants who'd been conceived about 42 weeks before scanning, although many of these had also been born prematurely.

Activation was observed in the same part of the brain in all three groups, showing that the BOLD blood oxygenation response measured by fMRI is present even early in life. However, in the pre-term infants, it was delayed: while in adults and typical infants it peaked about 6 seconds after stimulation, in the preterms it was more like 12 seconds. The size of the response was much smaller than in adults in both cases, though.

The authors say that the brain itself is probably not the source of the difference. Rather, they argue that in preterm infants, the blood supply to the brain takes longer to respond to the need for more oxygen.

Arichi T, Fagiolo G, Varela M, Melendez-Calderon A, Allievi A, Merchant N, Tusor N, Counsell SJ, Burdet E, Beckmann CF, and Edwards AD (2012). Development of BOLD signal Hemodynamic Responses in the Human Brain. NeuroImage PMID: 22776460

The Racist Brain?

Is the human brain... a racist?

There are some worrying indications that it could be. After all, the cerebrum is largely composed of so-called "white" matter, and the only black area is a little 'ghetto' at the bottom called, shockingly, the substantia nigra...!

Seriously though. There's a paper just out in Nature Neuroscience from Kubota et al that looks at The Neuroscience Of Race. It's a fine review as far as it goes, but to me at least, it really shows up the limits of contemporary neuroscience.

We are told that

A network of interacting brain regions is important in the unintentional, implicit expression of racial attitudes and its control. On the basis of the overlap in the neural circuitry of race, emotion and decision-making, we speculate as to how this emerging research might inform how we recognize and respond to variations in race and its influence on unintended race-based attitudes and decisions.

So there have been studies investigating which bits of American's brains activate in response to looking at photos of black people vs. white people. It emerges that "a network of interacting brain regions" light up. But so what?

Sure, the brain reacts differently to seeing people of different races. Of course it does - it reacts differently to everything, so long as we can perceive a difference; that's how we perceive a difference. And of course race, a deeply emotive issue in American politics and culture, activates 'emotional' parts of the brain - that's how it's emotive.

The included studies all scanned Americans and (presumably) mostly college students. Now most American college students are not active racists, and indeed I'd imagine that their emotional brains are more likely to be worrying self-referentially about racism than about the actual race of the stimuli. No-one seems to have scanned card-carrying members of the KKK. Furthermore, "race" in these studies almost always means "blackness". What about Latinos, Asians?

So what have we learned?

I don't think we've learned much about race. "Race" after all is a confused mixture of emotions, attitudes and beliefs. These differ greatly from person to person, and even the same individual may experience conflicting feelings in different contexts. Kubota et al note this and say that it may explain the mixed findings (black faces activate the amygdala more than white in some studies, not in others) but I'd have said that in this case, only inconsistent results are credible.

What does it tell us about the brain? I'd say not much. The authors weave a neat little narrative - in response to seeing black faces, the amygdala and other emotional areas activate as a negative emotional response; the ACC then detects this racist response and sees that it's unacceptable, and the DLPFC then suppresses it like a parent hurriedly interrupting a young child who's making a faux pas.

But all the elements of this story - the automatic, emotional amygdala, the supervisory DLPFC - are borrowed from other neuroscience studies so at best the race literature confirms these theories but it doesn't even really do that, because there are many other possible interpretations.

I'd say that we need to know much more in terms of the 'basic' neuroscience of emotion, attitudes and beliefs because we can tackle the hornet's nest of race in the brain.

Kubota JT, Banaji MR, and Phelps EA (2012). The neuroscience of race. Nature neuroscience, 15 (7), 940-8 PMID: 22735516

False Positive Neuroscience?

Recently, psychologists Joseph Simmons, Leif Nelson and Uri Simonsohn made waves when they published a provocative article called False-Positive Psychology

The paper's subtitle was "Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant". It explained how there are so many possible ways to gather and analyze the results of a (very simple) psychology experiment that even if there's nothing interesting really happening, it'll be possible to find some "significant" positive results purely by chance. Then you could publish those 'findings' and not mention all the other things you tried.

It's not a new argument, and the problem has been recognized for a long time as the "file drawer problem", "p-value fishing", "outcome reporting bias", and by many other names. But not much has been done to prevent it.

The problem's not just seen in psychology however, and I'm concerned that it's especially dangerous in modern neuroimaging research.

*

Let's assume a very simple fMRI experiment. The task is a facial emotion visual response. Volunteers are shown 30 second blocks of Neutral, Fearful and Happy faces during a standard functional EPI scanning. We also collect a standard structural MRI as required to analyze that data.

This is a minimalist study. Most imaging projects have more than one task, commonly two or three and maybe up to half a dozen, as part of one scan. If one task failed to show positive results, it need never be reported at all, so additional tasks would compound the problems here.

Our study is comparing two groups: people with depression, and healthy controls.

How many different ways could you analyze that data? How much flexibility is there?

*

First some ground rules. We'll stick to validated, optimal approaches. There are plenty of commonly used less favoured approaches, like using uncorrected thresholds (and then, which ones?) or voodoo stats, but let's assume we want to stick to 'best practice'.

As far as I can see, here's all the different things you could try. Please suggest more in the comments if you think I've missed any:

First off, general points:

• What's the sample size? Unless it's fixed in advance, data peeking - checking whether you've got a significant result after each scan, and stopping the study when you get one - gives you multiple bites at the cherry.
• Do you use parametric, or nonparametric analysis?

Now, what do you do with the data?

• Preprocessing
• How much smoothing?
• Do you reject subjects for "too much head movement"?  If so, what's too much?
• Straightforward whole-brain general linear model (GLM) analysis followed by a group comparison.
• What's the contrast of interest? You could make a case for Fear vs Neutral, Happy vs Neutral, Happy vs Fear, "Emotional" vs Neutral.
• Fixed effects or random effects group comparison?
• Do you reject outliers? If so, what's an 'outlier'?
• Do you consider all of the Fear, Happy and Neutral blocks to be equivalent, or do you model the first of each kind of block seperately? etc.
• "Region of Interest" (ROI) GLM analysis. Same options as above, plus:
• Which ROI(s)?
• How do you define a given ROI?
• Functional connectivity analysis
• Whole-brain analysis, or seed region analysis?
• If seed region, which region(s)?
•  Functional connectivity in response to which stimuli?
• Dynamic Causal Modelling?
• Lots and lots of options here.
• MVPA?
• Lots and lots of options here.

But remember we also got structural MRIs, and while they may have been intended to help analyze the functional data, you could also examine structural differences between groups. What method?

• Manual measurement of volume of certain regions.
• Which region(s)?
• VBM.
• Cortical morphometry.
• What measure? Thickness? Curvature...?

That's just the imaging data. You've almost certainly got some other data on these people as well, if only age and gender but maybe depression questionnaire scores, genetics, cognitive test performance...

• You could try and correlate every variable with every imaging measure discussed above. Plus:
• Do you only look for correlations in areas where there's a significant group difference (which would increase your chances of finding a correlation in those areas, as there'd be fewer multiple comparisons)?
• You could define subgroups based on these variables.

*

So even a very straightforward experiment could give rise to hundreds or thousands of possible analyses. 1 in 20 of these would give a statistically significant result at p=0.05 by chance alone, and even if you throw out half those for being "in the wrong direction" (and that's subjective in most cases) you've got plenty of false positives.

This problem is growing. As computation power continues to expand, running multiple analyses is cheaper and faster than ever, and new methods continue to be invented (DCM and MVPA were very rarely used even 5 years ago.)

I want to emphasize that I am not saying that all fMRI studies of this kind are in fact junk. My worry is that it's hard to be confident that any given published study is sound, given that papers are written only after all the data has been collected and analyzed.

I'll also point out that some imaging research, especially what might be called "pure" neuroscience investigating brain function per se rather than "clinical" studies looking at differences between groups, has many fewer variables to play with, but still quite a lot.

As to how to solve this problem, the one solution I believe would work in practice is to require pre-approval of study protocols.

Simmons JP, Nelson LD, and Simonsohn U (2011). False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological science, 22 (11), 1359-66 PMID: 22006061