X
Innovation

IBM is a step closer to developing accurate AI prediction model for Alzheimer's

In collaboration with Pfizer, Big Blue has developed an AI model using speech samples provided by the Framingham Heart Study.
Written by Aimee Chanthadavong, Contributor
human brain and artificial intelligence
Getty Images/iStockphoto

IBM has partnered with pharmaceutical giant Pfizer to design an artificial intelligence (AI) model to predict the eventual onset of the neurological disease seven years before symptoms appear.

Alzheimer's is currently incurable and is often diagnosed too late to prevent it from accelerating. Symptoms for the disease include the gradual degradation of memory, confusion, and difficulty in completing once-familiar daily tasks.   

Published in The Lancet eClinical Medicine, the researchers used small samples of language data from clinical verbal tests provided by the Framingham Heart Study, a long-term study that has been tracking the health of more than 5,000 people and their families since 1948, to train the AI models.

The AI model's ability was then verified against data samples from a group of healthy individuals who eventually did and did not develop the disease later in life. For example, if the AI model analysed a speech sample from a participant at the age of 65 and predicted they would develop Alzheimer's by the age of 85, researchers were then able to check records to determine if and when a diagnosis had actually occurred. 

Read more: How AI can help diagnose neurological disease (TechRepublic)

According to Big Blue, the outcome of this research was significantly better than predictions based on clinical scales, a prediction based on other available biomedical data from a patient, as that only had an accuracy rate of 59%. 

IBM added that unlike past studies, this new one focused on individuals that started to show symptoms or had a genetic history associated with the disease and only examined healthy individuals with no other risk factors.

"In partnership with our colleagues from Pfizer, we saw the potential to develop AI models which -- if continued to be trained on expanded, robust and diverse datasets -- could one day be used to develop methods to more accurately predict Alzheimer's disease within a large population, including individuals with no current indicators of the disease, no family history of the disease, or signs of cognitive decline," IBM said.

IBM said the ability to identify higher-risk patients could potentially lead to successful clinical trials for preventative therapies.

"Ultimately, we hope this research will take root and aid in the future development of a more simple, straightforward and easily accessible tool to help clinicians assess a patient's risk of Alzheimer's disease, through the analysis of speech and language in conjunction with a number of other facets of an individual's health and biometrics," the company said.

This latest research is part of IBM's ongoing research into Alzheimer's disease. In 2018, the tech giant introduced machine learning to the diagnostics field in the hopes that one day it could assist in the creation of stable and effective diagnostic tests for early-onset of the disease.

Related Coverage

Drug development in a year? IBM bets on AI, robotics to ramp up molecule research

IBM's new platform could contribute to ongoing research efforts at a time when discovering new drugs is at the heart of many projects.

IBM and Michael J. Fox Foundation develop machine learning model for Parkinson's

In hopes of helping clinicians pinpoint different stages of the disease as it advances.

IBM researchers tout potential key to reversing antibiotic resistance

Using laboratory automation methods, IBM researchers deployed unique synthesis methods to rapidly generate different potential polymers to be combined with therapeutics, allowing the testing of options for increasing antibiotic strength.

IBM tests the use of artificial intelligence for breast cancer screenings

The research has uncovered how AI can help radiologists improve the overall accuracy of mammogram screenings.

Editorial standards