AI brain scans detect early Alzheimer's with 90% success

A breakthrough in early diagnosis
26 March 2024

Interview with 

Tim Rittman, University of Cambridge


MRI scans


The best way to reduce the rising cost of Alzheimer’s disease is to heavily invest in research. And we think that one of the most critical targets in combating the condition is to find ways to diagnose the disease much earlier than we currently can to make treatments designed to slow or halt the progression as effective as possible. And that’s where Tim Rittman, at the University of Cambridge, comes in. He’s using artificial intelligence to detect the disease before symptoms set in...

Tim - There's one way we can try and get at that by looking at people who have say genetic dementias and we know we're going to get that in the future. So they'll have dementia passed down through the family and they carry a gene, which means they've got dementia later on. If we look at people in those groups, we can pick up subtle changes in the brain about 10 years before they develop symptoms. We can also look at the healthy population of people who go on to get dementia and we can pick up some subtle changes in their memory and thinking and general function about five to nine years before people develop symptoms or get a diagnosis.

Chris - How are you trying to do this?

Tim - So we've done this in a study called the UK Biobank and that's about half a million people across the country who volunteered to have tests and some of them had brain scans. So we took a computer algorithm which we'd trained to detect and sniff out Alzheimer's disease, if you like, by looking for these subtle changes in brain shrinkage. And we applied it to these, these healthy people who are around the age of 45 to 65, that sort of age group. So some of them might start to develop subtle signs of Alzheimer's disease. So we've got a big group of people who we know have Alzheimer's disease, so we can train our algorithm to sniff out Alzheimer's disease in those people that were confident about the diagnosis. And then we took it to this healthy group for about 40,000 people. We managed to pick up about 1300 who looked like they might have the start of Alzheimer's disease on their brain scan.

Chris - Can you not just use cognitive tests to find these people? Do we need to do additional, say brain scan testing on these people?

Tim - We have looked at the UK Biobank and those people who went on to get Alzheimer's disease. When you're looking at half a million people, we found these subtle changes in memory and thinking, but they were so tiny that you wouldn't be confident to say, well this person's definitely gonna go and get Alzheimer's disease just based on their cognitive test.

Chris - I suppose that's a reflection on the fact the brain is extremely good at compensating early on in the course of a lot of these degenerative diseases to overcome and surmount the problems that will later manifest.

Tim - Yeah, exactly that. Yeah. We know the brain changes the way it works and actually puts in a bit more effort early on to sort of cover up for the fact that there's early dementia going on.

Chris - Could we use this as a way to identify early people who might benefit from intervention either modifiable risk factors that we heard about earlier, smoking, drinking diet and so on, or also taking medications that might arrest the progression of Alzheimer's disease?

Tim - Yeah, I think it is. The study that we did was a sort of proof of concept, if you like to say, well this is possible to hunt out these people who might be at risk of future Alzheimer's disease. And I don't know if they're definitely going to get Alzheimer's disease, but they're certainly at high risk from it. So yeah, there's a lot of interest to try and pick out people like that for clinical trials to test out drugs that might stop or slow down Alzheimer's disease. I wouldn't be confident enough yet to say that these tests will definitely predict that someone's going to go and develop dementia. What we don't want is someone to have a brain scan and says, yep, you, you're going to get dementia because that will be the wrong message. But yeah, certainly finding people who are high risk, who might benefit most from those trials or interventions is really important.

Chris - I suppose one downside though is that we do need a brain scan, which is not a trivial thing to have for these patients. I recall David Cameron, former UK Prime Minister saying positive things about blood tests coming down the track. Is that an alternative?

Tim - Yeah, blood tests are definitely coming for, to support diagnosis of Alzheimer's disease. Probably in the next year or two we'll start to see them in memory clinics and yes, they'll certainly be easier. There's still a lot of question marks as to how accurate they are very early on. But more research is coming to try and increase their accuracy. But it might be that, you know, a combination of blood tests, brain scans and other tests might be the best approach in the end.

Chris - And what are the blood tests looking for? Because it's obvious with the brain scan, you're saying you can see subtle structural changes that are revolving in a person's brain, but what would a blood test detect?

Tim - So if you look at the brain of someone with Alzheimer's disease or another type of dementia under the microscope within the brain cells and around the brain cells, you see little clumps, little dots and those little dots and clumps are made up of these proteins. And the most sensitive blood tests so far are for a little protein called tau. And we've now got very sensitive assays which can pick up tiny, tiny, minuscule amounts of this tau, which has seeped into the bloodstream from the brain. And that seems to be the most sensitive test for Alzheimer's disease.

Chris - And in your view, if you were to take people whom you can identify with either blood tests or with brain scans as being at risk of Alzheimer's, what sort of a difference do you think we might be able to make to their outcome and to what sort of extent and over what sort of timescale? So how much earlier could we intervene and therefore what might be the overall benefit?

Tim - At the moment, you know, any answer to that is a guess. I think, we don't know, is the short answer, but I think we will be able to pick people up five to 10 years before they develop symptoms. That would be my hope. And if we can have a treatment which is safe and which is effective in those years, I think we could make a huge difference to people's lives. The challenge is that if we start treating people that early on, it's gonna take decades before we know the answers. But it's not just drug trial treatments either. It's also lifestyle interventions addressing things like smoking and blood pressure that we talked about earlier. So yeah, I think treatments and drugs are one thing and lifestyle and non-drug approaches might also be effective or part of the whole package.


Add a comment