An artificial intelligence system uses a dated graphic of two children stealing a box of cookies behind their mother’s back to detect the onset of Alzheimer’s disease.
The system uses sound samples of people describing the scene to predict the possibility of disease in healthy people.
It can predict the onset of disease seven years before doctors with an accuracy of 70%.
The Alzheimer’s Association UK said the research was encouraging.
“A timely diagnosis of dementia is critical to living well with the condition,” said Fiona Carragher, director of research at the Alzheimer’s Association.
“Unfortunately, it is often a long road that delays people with dementia from accessing appropriate treatment, support and guidance, yet current treatments are more effective earlier than they are received, and potential new treatments may also be more effective in the early stages, or even before,” she added Symptoms appear.
She believes that this is another “exciting step in the use of artificial intelligence and language to obtain early and more accurate diagnoses, although we need to see these approaches that have been further tested in larger and diverse groups of people.”
“We urgently need more research of this kind, given that Alzheimer’s disease causes changes in the brain up to 15 years before symptoms appear.”
Alzheimer’s disease damages memory and other cognitive abilities by destroying the connections between nerve cells in the brain.
In the United States alone, it is estimated that the number of infected people reaches 5.5 million, and some studies indicate that it is the third leading cause of death after heart disease and cancer.
The artificial intelligence model, developed by IBM Research and the pharmaceutical giant Pfizer, uses natural language processing to analyze short excerpts of speech taken from the “Cookie Theft” cognitive test.
The test, which has been used for many years to diagnose dementia and other cognitive diseases, asks people to describe what they see in the picture.
The artificial intelligence detected slight changes in the language, such as grammatical errors and different sentence structure, indicating cognitive decline.
The samples used by IBM came from the Framingham Heart Study, a US research project that included 5,000 people and their families, and has been going on since 1948.
Given the long-term nature of the study, the AI was able to analyze samples collected while the subjects were cognitively healthy. If the model analyzes a speech sample from a participant at the age of 65 and predicts that he will develop Alzheimer’s disease by the age of 85, this can be checked for accuracy.
“The main finding is that seven years before a clinical diagnosis, we can say with 70% accuracy that Alzheimer’s disease will develop in people,” said Ajay Ruyuru, Vice President of Healthcare Research.
He told the BBC that although there is no known cure for the disease at this time, knowing who might contract it could help with treatment.
“Even if you cannot change the course of the disease, you can prepare. You can manage your life better.”
Not every patient wants to know this, Ruyoro acknowledged.
“It has to be done through a strong ethical structure, and individuals must consent to the use of the tags, and if you choose to participate, you will find out. You will not be able to discover that unless consent is given,” he added.
But he hopes in the future, AI systems will help clinicians understand the role vital signs such as sound can play in diagnosis and predictive medicine.
“We believe that artificial intelligence is a way to enhance this process, it is just another indicator for a trained doctor, and it will not be artificial intelligence that gives the diagnosis.”