Mapping the human brain: interview with Emma Robinson


A team lead by American neuroanatomists M. Glasser and David C. Van Essen published a paper on Nature on the 20th of July this year detailing their latest exploits in mapping 180 areas of the brain. The map will largely improve precision for studies of the structure and task locations of the brain, its changes across individuals and in growing, aging and disease.

Emma Robinson

Dr Emma Robinson

Dr Emma Robinson from Imperial College London has kindly given an interview to KCW Today explaining how she came to work in neuroscience and the importance of this new paper. This interview follows up on the discussion of the KCW Today September print issue.

Tell us a little bit about yourself.

I’m Emma Robinson, and I’m a postdoctoral researcher at Imperial College London. Before that I worked  for a group specialising on functional magnetic resonance of the brain (FMRIB) based at Oxford University. That’s when we worked with the Human Connectome Project (HCP). The goal was to use MRI (magnetic resonance imaging) to generate high quality images of the human brain’s structure and function.

What made you choose Physics, and then Neuroimaging?

I did physics at undergraduate. When I came to the end of that, I looked for work on real world problems where I could help people.

I went and did a masters in general imaging and worked with neonatologists in imaging infants. I discovered that I really enjoyed the method side, creating tools that people can use to better understand the brain.

Since then I have been working in the computing department developing software to analyse brain imaging. I firmly believe it’s important to work in an interdisciplinary manner.

One of those analysis tools has been instrumental to this particular project.

Would you say the Human Connectome Project is a parallel of the Human Genome Project, but for neuroscience? What progress would we expect from a more complete knowledge of the human connectome?

Yes, it is inspired by the human genome (HGP). The difference is that the human connectome can be imaged over different scales.

I think the human genome would be equivalent in brain mapping terms to mapping all of the nerve cells and all of their connections in the brain, whilst what we are doing here is an approximation of regions of the brain and how they’re connected.

This will enable comparing these regions across different subjects in a more accurate way. It means we will be more sensitive to changes in different neurological conditions, understand different cognitive activities and be able to better model the brain.

I have the paper you were part of here: “A multi-modal parcellation of the human cerebral cortex.” Would you like to summarise it in a few words, and give us your opinion on its importance ?

I would say that this paper represents a paradigm shift on how we map and understand the brain.

There’s been an idea that the brain can be represented in this sense for a long time. But because of the vast scale of the brain where there’s over 100 billion brain cells and over 100 trillion connections, no has ever mapped all of those regions in any human brain.

It moves the field forward significantly by not only identifying 83 areas that we know, but 97 completely new areas that we didn’t know existed before.

This is really important. People performing neuroimaging studies will be able to compare those regions across different subjects. They will start looking for differences in populations for example, such as those with dementia and those with healthy subjects. Then we can narrow it down more successfully with what is causing those conditions – where that is happening in the brain, how is their connectivity involved in the works of that disease.

There is a large number of key findings in this paper that the HCP want to apply to Neuroimaging analysis. First of all I think it’s really important to do analysis by comparing data of the surface of the brain and second of all it’s important to look for correspondences between subjects through comparing brain function and not surface folding patterns.

The human brain has both a cerebral cortex, also called the neocortex, and the (evolutionary) older cerebellum, or little brain. Did your study also include information on the cerebellum ? How was that decision made?

Matthew Glasser has kindly answered this question by email.

“The cerebellum was included in the sense that differences in functional connectivity patterns across the surface included all of the brain’s grey matter, including the cerebellum.  We did not try to make a map of the cerebellum because mapping the cortex was quite an undertaking already.

It will be nice to map the cerebellum using similar methods, however the cerebellar cortex is much thinner than the cerebral cortex (what we mapped) and this makes doing the analyses properly harder.”


David Van Essen Left, Matthew Glasser right. Credit University of Washington

What important discoveries and technology have lead up to your current work?

This particular paper has pulled together a series of advances in medical imaging analysis on image processing that have been made over the course of the Human Connectome Project.

The HCP acquired very high quality imaging data and refined image processing pipelines in order to represent this imaging data on a cortical surface mesh.

As well as that I have contributed to this paper by providing a means with which we can compare imaging data across different subjects. That’s important when you are looking for regions that correspond to a particular brain activity. When the imaging data itself is quite noisy, we want to be able to identify that this pattern of activity appears on all subjects and be very confident that that’s real.

We also pulled together a classifier to be able to see it in all different subjects. It looks for those regions in a group, then propagating those regions to each subject and be able to predict those in a completely different group of subjects

The way they did that was by training a neural network algorithm to identify what it is about those regions that defines them. For the language region there would be language tasks that light up in the brain. In those algorithms they would be looking for patterns of activity in the new subject to identify those regions.

Were there any individuals with particularly different brain topology (map) found in your study?

One of the really interesting things about this paper was that when they were identifying these regions, some regions would appear in a different place in a subject relatively to other subjects. There’s a region called 55b, talked about extensively in the paper. This region changes its position relative to neighbouring regions in some subjects with relation to others.

That’s really important because in the past we’ve always assumed that all subjects have the same basic brain structure, that all their brain regions appear in the same form, in the same place. In this way this project will massively improve further studies’ ability to detect certain types of brain activity across the population.


White matter fiber architecture of the brain. Measured from diffusion spectral imaging (DSI). Not related to this study. (HCP)

There are records that Einstein had extra folds of white matter in his visual cortex, whilst nothing out of the ordinary was found with the brain of Carl Friedrich Gauss’s, one of the greatest mathematicians of all time. Would you like to comment on this?

Yes, it’s really interesting. Our feeling from the HCP is that the relationship between surface folding patterns and brain activity is not clear. There is some evidence to suggest that specialised functional regions in the brain do not always appear in the same position in a cortical fold in all subjects.

To an extent that says that you can’t tell a lot from the surface folding patterns from people’s brain function. But I think that as these particular functional regions or the function specialised regions in the brain, they are defined by their connectivity, and their connectivity may be related to how cortical patterns patterns fold.

It may happen that in time we will better understand the relationship between folding and function and it may well mean that there is some relationship between intelligence and folding as suggested with Einstein.

Tell us more about the classification. How was the number 180 for each hemisphere decided?

There is more likely to be a more fine grained parcellation of the brain, but the regions that they found were the ones they (Matthew and David) were most confident that were there.

This comes back to the fact that only part of the brain has ever been mapped before. They  did have a fairly confident idea of what they were expecting from these regions like in the visual cortex and in the motor cortex, that have been extensively mapped in post-mortem data in human and monkey brains.

On top of that they found a large number of new regions. That really comes down to their expert understanding of how the generic structure of these regions and quality of the imaging data they are using for identification.

Because there is evidence that regions aren’t appearing in the same place for some subjects and despite that this is some of the highest quality data that there’s ever been, there’s still noise in the data. In some areas it wasn’t possible to tell with absolute confidence if there was a regional boundary or noise.

3D MRI brain scan

Your current research focuses in neonatal and developing brain scans. Would you like to comment on any interesting changes on the brain at this stage and how it reflects on motor skills, behaviour and cognition?

If we could map the period of fetal and neonatal development it would have huge potential.

When you think about magnetic resonance imaging, its at a fairly low resolution. You are looking at the brain at the millimeter scale.You know that your nerve cells are much much smaller than that, on the micron (1000x smaller than millimeter) scale. Its a coarse approximation, we’re modelling what we think is going on and there’s room for error in that.

At the neonatal brain so much is changing that we can observe that change at the millimeter scale. In the adult it’s possible that we might not be able to detect all the difference that perhaps contributes to complex neurological diseases because they happen at a finer scale than we can actually observe.

Like the HCP we’re also collecting genetics data that we process at a later time. When the children get a bit older we get them to do cognitive tests so we can relate all of this to development.

Through that we will get much more information about how the brain is organised and how this differs between children that develop normally and those that perhaps have neurodevelopmental conditions.

The HCP is currently seeking healthy pregnant participants for their neonatal data scans. MRI scans are safe for mothers and their baby, with no known side effects. More information is at


Featured photo credit: Brain surface parcellation in 3D. Matthew Glasser, David van Essen, University of Washington

Gif: 3D MRI of the Human brain by Plasma Ben

About author