Inside young minds – unlocking teenage brains with MRI technology

Researchers are using MRI technology to help doctors start predicting the onset of schizophrenia and other psychiatric conditions, thanks to a collaboration between Cambridge University and University College London. The research, conducted under the umbrella of the Neuroscience in Psychiatry Network, saw 300 healthy patients aged 14–24 scanned over an 18-month to two-year period.

Led by head of psychiatry Ed Bullmore, the study found a link between early adult brain activity and schizophrenia risk. “We used MRI to see where we could find adolescent changes in the cortex, and then combined those results with gene expression data from the Allen Institute for Brain Sciences, so that we could work out which genes were strongly expressed in the brain regions that showed the greatest rates of change,” explains Bullmore.

The team showed that the genes that were strongly expressed in the brain regions associated with adolescent change were enriched for schizophrenia, which is how the mental illness came into the research, Bullmore adds, because the team hadn’t actually scanned any people who had been diagnosed with it.

“We were expecting there to be a change in the thickness of the cortex during adolescence: the cortex shrinks a little bit as you go through the process.

“What was novel was that shrinkage of cortical thickness was associated with increased myelination.”

Prior to this, it had been assumed that white matter was myelinated and the grey matter was not, but the scientists allowed for the idea that some of the cortex could be myelinated and then measured it, which showed that it also changed in adolescence in a way that was strongly coupled to the shrinkage of cortex.

“MRI is a good technique for measuring myelination, but we had never used it to measure myelination of the cortex,” Bullmore says.

The initial study is only the start of the process. It was done with one scan per subject, and followed up with some of the participants, although those results haven’t yet been analysed or reported. Bullmore says it’s difficult to predict what will happen over a long time with those brain network changes.

“The research has an important implication for adolescents. There isn’t much awareness that the brain is developing during adolescence. The general view is that behavioural changes and mood changes are mostly to do with sexual development and hormonal changes.

“The brain is going through a very active developmental phase, changing its structural and functional organisation. While some of these changes have been previously recognised, they haven’t been so tightly linked to myelination or to the development of the human brain network that’s so important for cognitive and other aspects of function.”

Stronger together

Three machines were used across three sites for the study: one at University College London, plus two machines based at Cambridge sites. They were all 3T Siemens Trios which meant they could do pilot work to demonstrate good reliability of data across sites by using the same manufacturer.

“People are often concerned about doing MRI studies over multiple sites, but it can certainly be done,” says Bullmore. “It’s easier if all the sites have the same field strengths – ideally the same manufacturer – and they have teams that are competent and willing to collaborate. If you’ve got those elements and you invest the time upfront in making sure you’re getting comparable and reliable data at each site, then it’s not an impossible challenge.”

However, the key was not just the technology, but also what the team was able to do with it. They used a sequence called multiparameter mapping, a “step advanced from regular MRI sequences,” according to Bullmore.

“It gave us some nice data that we could use to give estimates of cortical thickness and the myelination using magnetisation transfer sequences. We were able to do things in terms of analysing the data that hadn’t been done before.

“Having taken these good quality measurements of cortical thickness and cortical myelination, we did a lot of things to analyse the data and understand how changes in those measures were related to the brain network organisation and to brain gene expression. We also used a test-bed of 100 to develop the analysis methods and to iron out all the wrinkles in our data processing pipeline, and then applied the exact same data analysis methods to the other 200.

“We were able to replicate what we’d seen in the first 100 and that gives us confidence that what we are talking about here is likely to reproduce in future studies.”

MRI and psychiatry

Bullmore says the next challenge for psychiatry would be to move MRI away from being used exclusively as a research and diagnostic tool, and into the clinical arena.

“People with first-episode psychosis want to know what the future holds: am I going to get better? Am I going to be able to go back to university? Will I be able to live independently of my parents? At the moment, we don’t really have answers for them. The best answer amounts to ‘wait and see’.”

He believes researchers could do a better job of using imaging to understand where symptoms are coming from in terms of brain structure and in predicting outcome. “At the moment, we don’t use imaging to guide treatment or the expectations of patients and their families, and we could do that. The next thing after prediction would be prevention, or what I call ‘secondary prevention’.

“With schizophrenia, the brain develops abnormally over the course of postnatal life and then, in late adolescence/early adulthood, you get psychotic symptoms emerging. There’s imaging evidence to indicate that after the first episode of psychosis, about a third of patients will go on to have recurrent psychotic episodes and, in the worst cases, they will have chronic symptoms of schizophrenia. In those cases, you can see with imaging that there’s progressive change to the brain.

“In terms of treatment advances, if we could do a better job of predicting the outcomes, that would be great, but if we could intervene after the patient had presented clinically and offer treatment to minimise the risk of progressive brain changes, that would be exciting. It might reduce the chronic aspect of the disorder and prevent people becoming disabled. But to be able to prevent it, you have got to be able to predict it.”

Bullmore says that if treatment options existed, MRI could be used to predict future psychotic episodes at a much earlier age. “In terms of what is going on in the brain, nine to ten-year-olds may well already be on a different developmental path. If they’re destined to become schizophrenic, then there may already be changes in brain development and changes in brain organisation that we could pick up at ten years old.

“[However] doing that in clinical practice is a challenge, because of how many people you would have to screen, and you’d have to have a very low false positive rate because you don’t what to be telling 11 or 12-year-olds and their parents that they’re going to be schizophrenic unless you’re absolutely sure.

“Moreover, why would anyone want to know if there wasn’t a treatment for it? The idea that you could use MRI to predict schizophrenia in young children is technically achievable, but clinically it’s not a good idea until there’s something that you can offer in terms of changing the outcome. The secondary prevention idea, where you scan everyone once they’ve developed symptoms, is a bit more tractable.”

Future brain research

Cambridge University has installed a GE PET-MRI and hyperpolariser to continue its existing PET programme, and to start dynamic nuclear polarisation studies in humans, and while Bullmore describes the new technology as “exciting”, future MRI brain research may not centre on hardware.

Bullmore sees new types of machines as being able to help with different research streams: “We’ll probably use them for small, detailed studies rather than the larger-scale work.

“We should be able to get good precision on intracortical structure and intracortical myelination.”

Bullmore also expects it to give a higher signal-to-noise ratio, and superior spatial, time and spectral resolution, but doesn’t know what it will actually show that wasn’t already known in terms of how the brain is organised. He believes the progress in the field is not just about the hardware.

Many of the more recent advances in MRI have been either data analytic, in sequence development, or organisational, he says. The way they analyse the data has evolved and they are getting better at assembling large, high-quality MRI data.

“Scaling up MRI research so that you’ve got decent-sized samples that are more representative of populations, and are more likely to be generalisable and replicable, is important,” Bullmore says.

“There’s also been greater use of multisite data acquisition platforms, and there’s a lot more MRI data out there in the public domain than there was five or ten years ago.”

Bridge the gaps

The next challenge Bullmore sees is trying to get the advances that they’ve made in MRI research to translate to MRI in clinical practice. “The way MRI data is analysed in clinical practice is behind where we are in the research world,” he argues. “If we’d done our study with 300 kids and had analysed the data in the same way that data is analysed clinically, we wouldn’t have found out anything. If I were a funder, I would be saying to the MRI research community, ‘There are volumes of MRI research being published all the time, but where’s the clinical impact?’.”

Bullmore believes the cause could be gaps in radiological training that don’t include any significant emphasis on quantitative analysis. “It’s an informatics gap and what it requires is an infrastructure that can take standard clinical MRI scans and put them through the same kind of data analysis pipelines that are taken for granted in the research community [so that they] spit out quantitative results and provide a more complete account of an individual patient’s data and a way of comparing that patient to the healthy population in the same age range. That way, you’re able to make diagnostic comparisons more quantitatively and more objectively than at present.

“There’s an awful lot that we can do with the systems that are available. There’s a lot more we can find out from that technical base to achieve greater clinical impact.”

It’s a recurring theme with Bullmore: the potential of MRI is not necessarily in technology, but in the way that it is used.