Cambridge Festival: Speaker Spotlight

Dr Sam Nallaperuma-Herzberg is a Senior Research Associate at Cambridge University’s Department of Computer Science and Technology, and a Fellow in Artificial Intelligence (AI) for Science and Society at the Accelerate Programme for Scientific Discovery.

How does relaxation music affect our brains, and can it be tailored for therapy? New research from the BrainTwin project, part of the Accelerate Programme for Scientific Discovery, shows how AI and neuroscience can create sleep and relaxation treatments. At this event, Dr Sam Nallaperuma-Herzberg and her team will discuss their ongoing work and showcase a simple demo of early versions of the music therapy and brain signal analysis modules.

 

What inspired your interest in using Artificial Intelligence (AI) to treat brain disorders like insomnia?
I’m an insomnia sufferer myself. My interest in healthcare may be due to my family background with first-hand experience into the amazing world of healers. Clinicians and healthcare professionals do wonderful work, around the clock, caring for the fellow humans from cradle to the grave. And yet, they are under enormous pressure: the pandemic has created such huge backlogs for some treatments.

So when I first started working in AI, I began wondering if I could use my skills to help. At the time, I was modelling the use of AI in transport and communications infrastructure networks. These are complex networks, as are the ones in the brain. I realised that aspects of my work could be applied to brain networks.

It inspired my interest in neuroscience and I began thinking about how we could build digital twin models of the brain to understand it better. I developed a research proposal, was fortunate enough to receive funding, and now I’m leading the ‘Brain Twin’ research project in this area.

How can music help the brain?
Cognitive behavioural therapy is seen as the best of the NHS therapies offered for insomnia. But it’s expensive so patients can only have group, rather than individual, therapy (from a patient privacy point of view, this is not ideal). There are also some patients who are resistant to CBT, so they need different treatments.

Medication is another option, but it can cause side effects. The NHS doesn’t currently offer music therapy (though it can recommend it). Yet there is scientific evidence to show music can have a positive effect on insomnia. There are also clinical studies suggesting that the effects of music on people differ from person to person. These studies haven’t explored whether that is true specifically for insomnia – and that’s where the BrainTwin project comes in.  

How is artificial intelligence and neuroscience being used in the BrainTwin project to create music therapy?
We think music can be adapted to work for insomnia sufferers based on analysing the physiological signals from their brains. So we’re starting work on two studies to generate both subjective and objective data on that. We’ll be using a pre-trained AI model that identifies effective relaxation / sleep music.

Participants in the first study will be asked to listen to different types of music and tell us which makes them feel most relaxed. This will give us subjective feedback. (We already have volunteers for this study, but if anyone who comes along to our demo at the Cambridge Festival is interested, they too could take part.) This will give us subjective feedback. Then, to test whether people’s perceptions of what music works for them are correct, we’ll subsequently run a clinical trial and receive objective physiological feedback.

What else will the visitors be able experience at the Cambridge festival event?
Visitors will be invited to wear an EEG headset – or observe a team member wearing one. They can then hear sleep music generated by our models and see the live EEG waveforms, along with the sleepiness parameters. So they will basically see how AI generated therapeutic music makes them feel sleepy. 

Do you expect to see any difference between what music the participants say relaxes them and what their brainwaves tell you?
We know from our work with two sleep clinics that sometimes people’s perceptions of themselves do not always match with their physiological responses. Insomnia sufferers who are being monitored in a sleep clinic sometimes tell doctors in the morning that they had a very bad night when in fact the feedback from their physiological monitoring shows they actually got quite a few hours’ sleep. That’s why gathering the neurophysiological data as well as subjective feedback in our study will be so important.

How do you envision the future of music therapy in healthcare? Could AI-generated music become a mainstream solution for people struggling with sleep and relaxation issues?
I’m hoping that music therapy could be a tool to lessen the burden on the NHS by providing an easily accessible digital treatment for patients with mild insomnia. It could also be used for patients who are resistant to cognitive behavioural therapy. They could try it out and see if it works for them before the alternative, which is prescription medication that can have adverse side effects.