Harnessing ‘an enormous amount of statistical power’ to transform brain health diagnosis

Creyos is an online cognitive assessment platform built on more than three decades of research and a vast – and growing – trove of data.

The platform enables healthcare providers to evaluate working memory, attention, reasoning and executive function, potentially speeding up diagnoses and reducing waiting lists.

Health Tech World catches up with co-founder Professor Adrian Owen following Creyos’ Health Tech World Award win.

Hi Adrian. Where did the Creyos journey begin?

The project dates back 35 years to when I was a graduate student in 1988.

At the time, I had no intention of commercialising it of course. it was just part of my PhD.

Over time, as we published more research, people started asking to use our tests.

Back then, and this is going to really date me, I was mailing out floppy disks, then we progressed to USBs, until it became clear we needed to move everything online.

That’s when it became obvious that this could be a commercial product.

About 12 years ago, I co-founded Creyos with CEO Marc Lipton, transforming it into the platform it is today.

How do healthcare providers use the technology?

It varies depending on the provider’s needs.

We’re getting a lot of take up at the moment in predicting dementia among patients presenting at neurological clinics with memory or concentration problems.

When clinicians come across symptoms like these, what they want to know is, are they in the early stages of dementia or just experiencing the normal signs of ageing that we will all go through?

It’s really important to diagnose dementia early, not just so patients can access emerging treatments sooner, but also so clinicians can reassure patients who aren’t actually on the dementia trajectory.

I know that ADHD is another focus of yours, particularly given the ongoing waiting list issue in the UK and elsewhere.

This is a very important point, and I’m glad you expressed it in the nuanced way that you did.

Because while Creyos can’t provide a formal diagnosis, it can accurately predict the likelihood of an individual being diagnosed with ADHD, which is more or less the same thing.

If we can identify someone as highly likely to be diagnosed six months before they see a professional, this is really efficient for the system and improves waiting times for everyone.

Last time we spoke, you mentioned some long Covid research you were planning. How are you getting on with this?

We published a paper which showed some really interesting findings.

One was that there seems to be two types of people that have emerged with so-called long Covid – one group with mental health problems and one with cognitive problems.

The groups are actually quite distinct.

This is providing clues to what we mean by things like ‘brain fog’ in the context of long Covid.

And in terms of treatment, you’re going to treat cognitive problems differently to mental health issues.

It’s really exciting to be able to unpack those things using Creyos.

You’ve also conducted some research into gaming and cognitive function. What can you tell us about that?

We did this big participation study in collaboration with Manchester Science Festival.

When you review the available literature, no one really knows whether exercise is good for your cognition, beyond immediate effects like improving blood flow to the brain.

We wanted to find out whether exercise had any long-term benefits. So we compared two groups of people seemingly at different ends of the spectrum: exercisers and gamers.

It turned out that gaming was actually quite good for some aspects of cognitive function, whereas exercise didn’t have any effect beyond the mental health benefits.

And it seems that the more intensive, immersive games like Zelda where you’re having to navigate your way around an environment are much more beneficial to cognition than so-called brain training games.

So what’s next for Creyos?

I’ve always been cautious about claiming that we can ‘diagnose’ conditions with Creyos alone. But now I’m starting to think differently.

We’ve completed nearly 20 million tests now, and that gives us an enormous amount of statistical power.

With that amount of data, you can extract really nuanced information about the probability of a person with this particular profile going on to develop this disease.

Technically, this isn’t a diagnosis.

But if you can say with 90 per cent probability that this person will walk into a neurology clinic a year from now and be diagnosed, that’s pretty exciting.