Making thoughts visible. How artificial intelligence helps us to better understand brain activity

Left Marieke van Vugt and right Natasha Maurits.

Two studies, two types of bottlenecks

Natasha Maurits, Professor of Clinical Neuroengineering, focuses on changes in brain activity in cases of brain injury and ageing. For her research, she needs participants who are able to cooperate with measurements. Patients with brain injuries and older participants in particular are not always able to do so, which means that there are simply not enough people to participate. The problem therefore lies mainly in the availability of participants.

For Marieke van Vugt, associate professor of Cognitive Modelling, the bottleneck is different. She investigates how thoughts shift, wander or linger, for example when worrying or daydreaming. To study this properly, test subjects would have to spend a long time in the lab. In practice, this is hardly feasible. As a result, there is less information per person than her research requires.

Both are therefore looking for ways to gain more insight without placing too much of a burden on people.

Why artificial data can help

That is why the researchers are using artificial intelligence. Using so-called generative models (systems that learn to recognise patterns themselves), they can create simulated brain signals. You can think of these as imitations of the wave movements that appear in real brain measurements. This artificial data does not replace real measurements, but it can supplement what is missing. This makes it possible to investigate questions for which there would otherwise be too little material.

First understand, then apply

The use of AI does require caution. ‘We want to know exactly how such a model arrives at its results,’ says Maurits. ‘Only then can you use it responsibly.’ The researchers are therefore trying to understand the underlying logic of the models: why does a certain pattern arise and when can you trust the simulated data?

More insight into thought processes

For Van Vugt, this approach ties in with her research into the dynamics of thinking. She wants to understand how someone gets stuck in a thought, for example when worrying, and whether you can break that pattern. Together with her colleague Marie-José van Tol, she is looking at how interventions such as mindfulness and positive fantasising influence these thought processes. ‘If we learn to interpret brain activity better, we may be able to see early on when someone gets stuck in their thoughts,’ she says.

Cautious optimism

The research is still in its infancy, but the initial results are promising. The researchers have developed artificial brain signals that closely resemble real patterns. The next step is to link these to processes such as attention, memory and moments of distraction.

‘Ultimately, I would like to understand how brain activity changes with ageing or after brain injury,’ says Maurits. ‘AI can reveal patterns that would otherwise remain hidden. But we need to know exactly what the model does before we draw any conclusions.’

A broader significance

Because brain research is often limited by a lack of measurable data, this approach could offer new opportunities worldwide for research into brain injury and dementia, as well as for a better understanding of everyday thought processes.

Van Vugt: ‘The goal is not to replace the brain, but to understand it better. In this way, we are gradually building more insight into the most complex system we know: our own brain.’

This research is part of the Jantina Tammes School's M20 programme and is funded by the Ubbo Emmius Fund.

Would you like to donate? Become a friend or supporter of the Ubbo Emmius Fund and enable pioneering scientific research at the University of Groningen.

Text: Djoeke Bakker

Image: Reyer Boxem