(NewsNation) — Scientists at Cambridge University have developed an artificial intelligence tool that is better at predicting the progression of Alzheimer’s disease than current testing standards.

Researchers used data from cognitive tests and structural MRIs from 400 people in the U.S. They tested the model for accuracy on 600 people in the U.S. as well as 900 patients who were part of longitudinal studies in the UK and Singapore.

The AI model was able to assess whether patients had stable, mild cognitive impairment or were expected to progress to Alzheimer’s disease within three years. The model correctly identified 82% of those who would go on to develop Alzheimer’s and 81% of those who did not.

That’s critical because early detection of dementia can help ensure patients get treatment early, when it is more likely to be effective. Under current methods, up to a third of patients are misdiagnosed and others are diagnosed too late for treatment to significantly slow the progression of the disease.

Published in eClinicalMedicine, the AI model was three times more accurate than current clinical assessments. It also allowed researchers to group patients into three groups, those who were stable and those who would develop Alzheimer’s slowly or more quickly.

Of those studied, roughly 50% were found to have stable symptoms of memory loss, while about 35% would develop Alzheimer’s slowly and around 15% would see quick progression to Alzheimer’s disease.

Researchers noted the 50% found unlikely to progress to Alzheimer’s could be helped under the new model by considering other causes for memory impairment, such as depression. It could also help those likely to develop Alzheimer’s get treatment more quickly and alert doctors to more closely monitor those who are likely to have a fast disease progression.

The data was validated against the 900 patients from the UK and Singapore who had been studied over time, some up to six years, which allowed scientists to compare predicted disease progression from early tests with the actual experiences of those patients over time.

The team that conducted the study said that the AI model is one that could be introduced into clinical practice quickly.