MRI brain scans can help predict whether a person will develop dementia in the next three years, before the symptoms of the disorder appear, scientists have found.

In a study, researchers from Washington University and University of California San Francisco in the US used magnetic resonance imaging brain scans to predict dementia with 89 per cent accuracy.

“Right now it’s hard to say whether an older person with normal cognition or mild cognitive impairment is likely to develop dementia,” said Cyrus A Raji, an assistant professor at Washington University.

“We showed that a single MRI scan can predict dementia on average 2.6 years before memory loss is clinically detectable, which could help doctors advise and care for their patients,” said Raji.

Although there are no drugs available yet to prevent or delay the onset of Alzheimer’s disease, identifying those at high risk of developing dementia within the next few years could still be beneficial, the researchers said.

People could make decisions on their financial and living arrangements while they are still in full control of their faculties.

They used a technique called diffusion tensor imaging to assess the health of the brain’s white matter, which encompasses the cables that enable different parts of the brain to talk to one another.

Using information from the Alzheimer’s Disease Neuroimaging Initiative–a multisite collaboration that pools data, funding and expertise to improve clinical trials for Alzheimer’s disease–researchers identified 10 people whose cognitive skills declined over a two-year period and matched them by age and sex with 10 people whose thinking skills held steady.

The average age of people in both groups was 73. Then, the researchers analysed diffusion tensor MRI scans taken just before the two-year period for all 20 people.

The researchers found that people who went on to experience cognitive decline had significantly more signs of damage to their white matter.

The researchers repeated their analysis in a separate sample of 61 people, using a more refined measure of white matter integrity.

With this new analysis, they were able to predict cognitive decline with 89 per cent accuracy when looking at the whole brain. When the researchers focused on specific parts of the brain most likely to show damage, the accuracy rose to 95 per cent.