New AI device can significantly improve Alzheimer’s diagnosis after breakthrough in understanding causes

Scientists have developed an AI device that can diagnose Alzheimer’s much more accurately than the current “gold standard” test – and hope it could start to be used on the NHS as early as next year.

The new tool is able to accurately identify people with Alzheimer’s 83 per cent of the time, compared to 67 per cent for the best existing tests.

It was made possible after a breakthrough in understanding about the “amyloid” protein plaques that build up in the brain and trigger the disease – as well as indicating its presence in diagnostic tests.

The developers of the device discovered that it was not just the number of these plaques that was important but their positioning – a finding that could help researchers develop effective new treatments as well as improve diagnosis further down the line.

Alzheimer’s is a notoriously difficult disease to diagnose because there are lots of causes and symptoms.

But this new device “boosts the diagnostic power of a PET scan”, said Ashwin Venkataraman, of Imperial College London.

PET, positron emission tomography, scans are currently the most accurate diagnostic tool for Alzheimer’s and work by analyzing the build-up of plaques in the brain.

“I think it is a breakthrough, it is important. The positioning of amyloid plaques is crucial to be able to say ‘it’s Alzheimer’s’, or ‘it’s not Alzheimer’s’ – and we’re showing that for the first time with this method [PET scans],” he told I.

Dr Venkataraman hopes to do more research this year to further validate his findings, using data from the NHS or specialist memory clinics.

And in a best case scenario, he said his diagnostic tool could start to be used on the health service “in about a year”.

However, he cautions that more work is still needed to confirm the potential of the new device with no guarantee of success.

Sian Gregory, Research Information Manager at Alzheimer’s Society, which funded the research, said: “This is an important finding as in the future people with Alzheimer’s disease will be able to get quicker, more accurate diagnoses – the key to unlocking the right treatments and support.”

“PET scans are often used to diagnose Alzheimer’s disease, but can sometimes be difficult to interpret as the amyloid plaques they identify can also be seen in the brains of people who don’t have symptoms of the disease,” she said.

Experts not involved in the research welcomed the findings.

“This study shows the potential of artificial intelligence to diagnose Alzheimer’s disease more accurately. Methods like this have great promise to improve both diagnosis and clinical care, said John O’Brien, professor of old age psychiatry at Cambridge University and leader of a research program on dementia diagnosis at the National Institute for Health Research government agency.

“Getting an accurate diagnosis is of great importance for people with dementia and their families. It allows for proper management and is often a key gateway to accessing other services and support,” he added.

The study used machine learning to analyze PET scans for proteins in the brains of 758 patients with Alzheimer’s, and tested how they link to the patient’s measures of forgetfulness.

The research is detailed in the journal Alzheimer’s Research & Therapy. The Invicro clinical imaging center in London was also involved in the study.

Despite a current accuracy of just 67 per cent, PET scans are already the most accurate diagnostic test for Alzheimer’s, although there is considerable room for improvements.

Other methods include cognitive testing and CT and MRI scans.