Jul 06 2017

The Brain’s Wiring

connectome5Cardiff University has released its latest scans of the wiring of the human brain. This now adds to a similar project in the US funded by the NIH.

The result is a stunning image of all the axons in the human brain – the wires that conduct signals and form networks and connections to other parts of the body. In addition to these wires there are also the neurons, which are the cell bodies of which the axons are part, and the glia, which are other cells in the brain that serve support functions but also play a role in modulating neuronal function.

These wiring images are part of various connectome projects – attempts at fully mapping all the connections within the human brain and their functions. These latest images from Cardiff are the result of the Siemens 3 Tesla Connectome MRI system. Essentially we are seeing the result of advances in both hardware and software technology.

The MRI scanners themselves are on the more powerful side. The “3 Tesla” refers to the power of the magnet. A typical hospital MRI scanner operates at 1.5 Tesla or 3 Tesla. There is a 10.5 Tesla MRI unit at the University of Minnesota which is used for the NIH connectome project. So 3 Tesla is a powerful MRI scanner, but no more so than the more powerful typical clinical MRI scanners.

The power of the magnet, however, is not the only measure of the abilities of an MRI scanner. The Cardiff unit is “specially adapted” to have very high resolution. They can image fibers in the brain 1/50th the width of a human hair. There is only one other unit with this resolution in the world, at Harvard University.

In addition to high-resolution anatomical imaging, the Cardiff and NIH projects use function MRI scanners to image the axons in action. This information is supplemented with magnetoencephalography (MEG) and EEG scanning, which are also functional scans looking at brain activity. The goal is to combine anatomical and functional data to create the connectome – a map of the functional networks within the brain.

The resulting images are stunning. It reminds me of the Saturn V rocket in a way. In response to the moon hoax nonsense, one NASA engineer commented that – looking at the Saturn V rocket’s size, and doing even some basic calculations, it’s clear that rocket was designed to go beyond Earth orbit. It was large enough not only to reach the Moon, but theoretically even Mars. If we didn’t really go to the moon, why was the rocket so big?

I feel the same way looking at the emerging images of the connectome. If you look at all the connections in the brain, you have to wonder what they are for. All those networks and all that processing power is needed to do all the things that the brain does, including producing consciousness. I know that won’t convince neuroscience deniers, but for scientists it is amazing.

One comment on reporting about these images. The BBC writes:

Doctors hope it will help increase understanding of a range of neurological disorders and could be used instead of invasive biopsies.


The scanner is being used for research into many neurological conditions including MS, schizophrenia, dementia and epilepsy.

I know it is obligatory for any basic science reporting in the media to make a connection to some concrete application. It is, of course, a reasonable question to ask – how can we exploit this new technology or finding for something specific? Unfortunately, that is often where speculation goes off the rails, or at least is not put into perspective.

I do have great hopes for the potential of the connectome project, but it won’t lead directly to cures for MS (multiple sclerosis) or dementia, and it may not even contribute to their diagnosis. First let’s address research into understanding these diseases.

Some brain disorders, like MS and Alzheimer’s disease (AD), are pathological diseases. MS is caused by inflammation in the brain and spinal cord which causes direct damage to cells and axons. In AD cells in the brain slowly die off for still unclear reasons, although we have identified many pathological changes that are involved in the progression of the disease.

In other words, these are pathological diseases of the cells in the brain. They are not the result of disordered connections within the brain. Brain connections are disrupted secondary to the pathology, so they are the result of and not the cause of the disease.

This kind of imaging, therefore, might help us understand the consequences of diseases like MS and AD. I don’t think, however, they will help us understand the pathology, and therefore is unlikely to lead to effective treatments (which have to address the underlying pathology).

Schizophrenia is different. That is a disorder of brain wiring, and not necessarily pathological. Brain cells are fine, they are just networked to each other in a way that produces the signs and symptoms we define as schizophrenia. Connectome projects, therefore, will likely be much more useful to understanding disorders like schizophrenia which are the result of disordered connections.

What about using this technology as a diagnostic tool. That is possible, but not as straightforward as reporting makes it seem. In order to be useful clinically, any examination tool must give us specific clinical information that we can act upon. It has to distinguish different related diagnoses, or make a diagnosis earlier than we can by other methods, or predict outcome or who will respond to specific treatments. Just knowing in more detail exactly how a brain is misfiring won’t necessarily lead to any change in management.

I do not think it is likely that connectome scans will be useful in diseases like MS. I don’t see how this information will affect management. Further, I don’t think they will replace biopsies (to be clear, we don’t routinely use biopsies to diagnosis MS). Biopsies are about finding pathology, and as I said above, imaging the connections is not about cell pathology. They are simply different kinds of information.

The only clinical role I can envision for AD is making a diagnosis earlier than can be made clinically. Sometimes patients present with very mild symptoms of dementia, before they are advanced enough to confirm the diagnosis of AD or a similar dementia. Such a scan might show early connection loss and confirm the diagnosis a year or so before it can otherwise be confirmed. I don’t know if this will be the case, but it’s plausible.

Whether or not this is clinically useful is a separate question. It will be helpful for prognosis, which will help people plan for their future. It may not otherwise have a dramatic affect on management, but that may change as new treatments become available.

Overall I think the direct clinical applications are likely to be limited, but may find utility in specific situations. The research applications, however, may be massive, but they will not apply to pathological disease. Rather they will mostly apply to our understanding of how the healthy brain works, and in understanding disorders which are essentially problems with the brain’s connections.

12 responses so far

12 Responses to “The Brain’s Wiring”

  1. TheTentacleson 06 Jul 2017 at 8:54 am

    The images are stunning, but the article itself is so basic and tedious; I wish they made this available directly at the centre’s site, along with some more technical information (the centre website is pretty useless).

    These kinds of multi-DTI datasets will be really important as “scaffolding” for the functional data obtained via fMRI, EEG, MEG, and ECoG. Accurately correlating across these data sets is still a non-trivial analytical challenge however… Given large inter-subject variability, what constitutes pathological connectivity is also a challenge.

  2. Nidwinon 06 Jul 2017 at 9:24 am

    Impressive even if it looks a bit like some alien bush from planet x,y,z.

    I although got the feeling, watching the vid and when they zoomed in, that it looks very complex but also fragile making we wonder the wisdom of some sports or excesses and abuses in our lifes.

    As you mentioned Dr Novella I would too prefer that science and scientific journalism refrain from throwing in possible future applications concerning the subject of findings and just keep it to the sharing of result and information from the study. But that’s just me of course.

  3. edamameon 06 Jul 2017 at 11:07 am

    A couple of things. Ooh pretty fireworks fourth of July brain!

    Clinical optimism
    I’m much more optimistic about clinical applicability. I would be surprised if DTI didn’t become extremely important in about 50 years or so, once the connective basis of some disorders becomes better understood, stronger magnets are routine, and analysis is faster with high-speed computers, so getting someone’s connectome is as routine as getting a CT scan, and we have millions of baseline connectomes. It is impossible to predict specifics right now, obviously.

    However, I’m surprised about Dr Novella’s skepticism about how useful this could be in the study and diagnosis of MS, which is precisely a disorder of connectivity and myelination: DTI shows us where myelinated bundles of axons are–that’s exactly what it does, so it can reveal demyelination and axon loss. A connectome scan could be one dimension in the multidimensional space of diagnostic criteria, potentially helping to uniquely specify MS. I would defer to Dr Novella about the benefits of early diagnosis of MS, but I assume it is better to know earlier than later. The use of DTI in the study of MS is not mere speculation but an active research area with hundreds of studies: https://www.ncbi.nlm.nih.gov/pubmed/20882528. It is very promising and already seems to be pushing in useful directions.

    Psychiatry (for instance) is still largely correlation based voodoo. I’d be surprised if some diseases (or subtypes) didn’t partition neatly based on connectivity data (e.g., autism). But it could be that DTI isn’t high enough resolution to get the partitions right, and we’d need to do large-scale histology on tissue. Which I’m pretty sure people wouldn’t want. 🙂 But there are other ways to get functional connectivity than just DTI and histology (e.g., granger causality analysis of EEG or even LFP data — if it comes to that). So if it turns out to be crucial, and the disorder is bad enough, connectivity will still end up being a useful diagnostic tool.

    Beware that DTI only reveals large tracts (e.g., corpus callosum). There is a reason you don’t see local corticocortical connections in these pretty pictures. It isn’t because such connections aren’t there, or are unimportant. They are not bundled together into tracts, and such local axons are often unmyelinated! Also, dendrites are completely left out of these pictures, and dendritic morphology is not some functionally inert feature of a neuron (https://www.ncbi.nlm.nih.gov/pubmed/8684467).

    Also I was surprised when I worked with one of these diffusion tensor datasets how many parameters there were to play with. That is, it wasn’t as simple as just taking the dataset and getting out connectivity. It wasn’t plug and play. E.g. you have to set a threshold above which something will count as a ‘connection’. When you lower that threshold you get all sorts of spurious connections (false positives). So there is an art to this, just as there is to fMRI analysis. I had to enlist the help of an expert to guide me through the process.

    I’m not saying that invalidates it. I’m just pointing out that these pretty pictures you get have billions of false negatives, some false positives, and you are just seeing the major myelinated tracts. MRI will never replace old school neuroanatomy, the gold standard being electron scanning microscopy. But it isn’t like there is a conflict: as Dr Novella said the brain has processes at multiple spatial (and temporal) scales, and MRI is pretty helpful for revealing certain features at certain resolutions.

  4. edamameon 06 Jul 2017 at 11:14 am

    I should be clear when I wrote they have “billions of false negatives” I was being provocative not literal. The correct thing to say is that they don’t intend to reveal small local connectivity patterns, but only large-ish myelinated bundles of axons (I think on the order of a millimeter or so in diameter but I am not an expert on this frankly so I may be off). So to the extent that they aren’t missing those intended bundles of axons, then they don’t have false negatives.

  5. Steven Novellaon 06 Jul 2017 at 11:34 am

    I did not say that DTI is not useful in MS. It clearly is. It is already very useful for research. It also has significant implication for clinical management.

    The question is, will higher resolution full connectome data add any clinical utility. I don’t think it will be a game-changer, and may have limited or no direct clinical role. The question is – will it be useful (more useful than easier or cheaper methods) in determining management?

    Early diagnosis is useful, but mainly in allowing early treatment. But, we already have thresholds for who benefits from treatment. In other words, patients who are too mild or too early in their disease don’t get treated. So, pushing back the diagnosis to even more early or mild won’t change anything.

    Of course, this can change if we develop new treatments that are beneficial earlier.

  6. TheTentacleson 06 Jul 2017 at 11:34 am

    edamame: interesting point regarding the dimensionality of the analysis. Is there a “standard” analysis package?

    Regarding dendrites (which as you say are invisible to DTI), here is another very interesting take on their importance, regarding the apical dendrites of cortical cells: http://dx.doi.org/10.1093/nc/niw015 — dendrites and local recurrent connections are measurable using other connectomic analysis methods but often only in postmortem tissue.

    Note: DTI can also be used as an in-vivo experimental tool for better targeting of multi-areal recordings; for example this great paper by Saalman: https://doi.org/10.1126/science.1223082

  7. edamameon 06 Jul 2017 at 12:10 pm

    Tentacles: I am not up to speed with the standard packages in the field, frankly. I’ve only analyzed DTI data once, and with the help of a guide who really knew his stuff. One package that a lot of people seem to use is called ‘FSL’: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki

    But frankly I have no idea how good it is and I can’t vouch for it. My wheelhouse is electrophysiology and behavior.

  8. Lightnotheaton 06 Jul 2017 at 6:03 pm

    “All those networks and all that processing power is needed to do all the things that the brain does, including producing consciousness. I know that won’t convince neuroscience deniers, but for scientists it is amazing.”

    Anticipating that Ian Wardell et al will respond to this that actually a “courageous, enlightened, cutting-edge” group of scientists are against the “materialist” view that consciousness is reasonably explained as solely a product of the brain, and that this avant-garde view will eventually be embraced by the mainstream, I again ask, what evidence do we have of how large this group is, and whether it is growing?

    Because the counter argument is that there is always a fringe group opposed to the mainstream, without that meaning necessarily that that fringe group are “ahead” of the rest. And of course we can cite many examples of the fringe view dying out. But I think if you could show that the group is not that small, and steadily growing, that would be evidence of denialism going on in the mainstream view. Science makes assumptions and has ideological elements like any other world view, and denialism can occur. As I’ve said before, some of the objections to psi evidence can sound more than a little denialist.

    I’ve done some research but haven’t yet found any evidence, from surveys and such, of how large the group of Wardell-approved scientists is, and whether it’s growing. Have any of you guys looked into this?

  9. hardnoseon 06 Jul 2017 at 7:23 pm

    I love the picture. I think this technology will help scientists get more confused about how the brain works. It will be fun to see.

  10. Sylakon 06 Jul 2017 at 7:57 pm

    Because, of course, you know how it works. It’s the only way you can know they will be confused. Please, enlight us, oh great savant, about the wonder of your knowledge.

  11. bachfiendon 06 Jul 2017 at 9:14 pm

    The picture, of course, won’t confuse scientists on how the brain works. It’s a tool they’re been trained to use. Scientists prefer to use tools instead of subjective perception because they’ve been led astray by it in the past, such as with the non-existent N-waves.

    Hardnose of course prefers subjective perception, since it makes it easier for him to accept all sorts of nonsense.

    It’s laymen, such as hardnose, who are misled by pretty pictures, to the extent that brain scans are generally prohibited in courts of law, in case they over-impress juries.

    A book recommendation: ‘the Punisher’s Brain: the Evolution of Judge and Jury’ by Morris B Hoffman. Hoffman is actually a judge.

  12. Nidwinon 07 Jul 2017 at 2:47 am


    I disagree with you as I think it’s one “small” step closer to actually help us understand how the brain works and how it connects and relates to other parts of the body. It will take time to improve software, hardware, data models (testing and application) but as always when IT is involved this can go etremely fast if wanted.

    Confusion is part of being in STEM but it doesn’t stay long, esepcially when there’s something new and shiny to start to dig into, to validate and find an or multiple applications for.

Trackback URI | Comments RSS

Leave a Reply

You must be logged in to post a comment.