Line drawing used by AI to detect the onset of Alzheimer’s disease

An artificial intelligence (AI) system is utilising a black and white line drawing of two children sneaking into a cookie jar behind their mother’s back to identify the onset of Alzheimer’s disease.

For many years, the Cookie Theft cognitive test has been used to diagnose dementia and other cognitive disorders.

The AI system, which was formed by the multinational pharmaceutical corporation Pfizer and IBM Research, utilises natural language processing to examine small extracts of the discourse of the participants describing the photo, to predict whether healthy people may get the disease.

It spots subtle differences such as grammatical and sentence structure errors.

The discourse samples came from The Framingham Heart Study, which has examined 5,000 people and their families since 1948.

Due to the extended period of the study, the AI system could examine past samples of the participants, while they were not experiencing cognitive struggles. Therefore, the AI was able to check the accuracy of these past diagnosis predictions.

Researchers state it could predict onset with 70 per cent accuracy, and seven years earlier than human doctors.

Director of Research at the Alzheimer’s Society, Fiona Carragher, states that “this is an exciting further step in the use of artificial intelligence and language to get much earlier and more accurate diagnoses. Though we need to see these methods tested further in larger, diverse groups of people.”

Currently, there is no known cure for Alzheimer’s disease, which damages the connections between a person’s nerve cells in the brain, and alters their memory and other cognitive abilities.

IBM’s vice-president of Healthcare Research, Ajay Royyuru, says “he hopes that AI systems can help doctors understand the role bio-markers such voice can play in diagnosis and predictive medicine.”

Share this article:Email this to someone
email
Share on Facebook
Facebook
Tweet about this on Twitter
Twitter
Share on LinkedIn
Linkedin