healthHealth and Medicine

AI Tool Can Diagnose Dementia In Just One Day (Years Before Symptoms Even Develop)


Dr. Katie Spalding

Katie has a PhD in maths, specializing in the intersection of dynamical systems and number theory.

Freelance Writer

Brain scan

Alzheimer's disease, the most common form of dementia, as seen in an MRI scan. Image credit: Atthapon Raksthaput/

There are around 50 million people in the world today with some form of dementia, and that number keeps going up. And yet it’s still relatively poorly understood: most cases lack a definitive cause, and diagnosis can be a long and difficult process.

But it might be about to get a little better. Scientists from the University of Cambridge are trialing a new machine-learning algorithm that can diagnose dementia in a single day – and it’s able to spot the disease years before any symptoms develop.


“Traditionally, when we look at patient scans we are looking for patterns to be able to help us exclude things like strokes and brain tumors,” explained trial lead Timothy Rittman. “The computer can do this much more comprehensively than any human.”

So far around 80 patients have taken part in the trial, which is being carried out at Addenbrooke's Hospital in Cambridge, England, but that number is expected to rise to around 500 in the first year. Should the trial be a success, the algorithm could be rolled out to patients across the UK – a move with the NHS says could save the service up to half a billion pounds over the next five years.

The algorithm works by comparing patient brain scans against those of thousands of dementia patients, which it uses in combination with standard memory tests to train itself to spot the disease. It’s already shown itself to be capable of identifying even the subtlest tell-tale signs of the condition – signs that even expert neurologists might miss – and match them to patient outcomes from the database.

That’s very good news for patients. Not only does it make a diagnosis much quicker and less invasive, it spares them months or even years of disease progression. Potentially, it could mean avoiding symptoms completely.


“If we intervene early, the treatments can kick in early,” Zoe Kourtzi, who developed the algorithm, told the BBC. “[We can] slow down the progression of the disease and at the same time avoid more damage … it's likely that symptoms occur much later in life or may never occur.”

No two cases of dementia are exactly the same, which can make living with the disease even harder. Some patients can stay relatively stable for a long time, for example, while others deteriorate rapidly, and patients and families are often left feeling helpless in the face of that uncertainty. But by comparing scans with previous patient outcomes, the new algorithm doesn’t just diagnose dementia – it can also tell the newly-diagnosed what they can expect to face.

“[It is a] fantastic development,” Rittman told the BBC.

“These set of diseases are really devastating for people … when I am delivering this information to a patient, anything I can do to be more confident about the diagnosis, to give them more information about the likely progression of the disease to help them plan their lives is a great thing to be able to do.”


Receive our biggest science stories to your inbox weekly!


healthHealth and Medicine
  • tag
  • dementia,

  • algorithm,

  • machine learning