A new AI model is able to predict future onset of Alzheimer's disease around 7 years in advance of diagnosis using short speech tests, according to a new study published in The Lancet eClinicalMedicine.
The big picture: There's still no treatment for Alzheimer's, meaning that there could be limited real-world demand by patients for such a tool today. But it could also be valuable for recruiting patients for clinical trials for potential treatments.
Details: The model, developed by IBM Research and Pfizer, analyzed speech samples provided by the Framingham Heart Study, a long-term study that has been tracking thousands of people since 1948.
- The samples were collected before the subjects began showing signs of cognitive impairment. Researchers then trained AI algorithms to correctly predict the eventual onset of Alzheimer’s disease in healthy participants.
- “The production of speech is an intensely cognitive task. There is no preformed train of thought or train of words that is sitting somewhere in your head, and all you have to do is pull it out," IBM's Ajay Royyuru, vice president of Healthcare Research, told Axios.
Why it matters: “It's not yet part of any clinical assessment that anybody does, but it could become part of what you do with some frequency," Royyuru said.
- He also said that this kind of tool could be used to recruit patients who are the right stage of the disease for clinical trials, which could help increase the odds of finding an Alzheimer's drug that works.
- "Early interventions can only be effectively tested and implemented if the population that stands to benefit can be identified," the study's authors write.
The bottom line: An Alzheimer's treatment has so far proved to be frustratingly elusive, but new technology could help eventually find one and identify patients who would most benefit from early intervention.