Artificial Intelligence Predicts Alzheimer's Through Your Voice

6 Years Before It Happens

Scientists from Boston University have developed a new artificial intelligence algorithm that analyzes the speech patterns of people with mild cognitive impairment to predict the likelihood of developing Alzheimer's disease.

This algorithm can predict the progression of the condition from mild cognitive impairment to Alzheimer's disease within six years with an accuracy of up to 78.5%, according to the "Science Alert" website.

Detecting cognitive impairment

The scientific team is continuing its research published in June, by training a model using audio recordings of more than 1,000 people to accurately detect cognitive impairment.

The new algorithm was trained on audio transcripts recorded by 166 individuals with mild cognitive impairment (MCI) between the ages of 63 and 97.

Since the team knew in advance who would develop Alzheimer’s, a machine learning approach was used to find markers in the audio transcripts that linked the 90 people who had experienced cognitive decline to Alzheimer’s.

Predicting Alzheimer’s

After training, the algorithm could be used in reverse to predict the risk of developing Alzheimer’s from previously unprocessed audio transcripts.

Additional important factors, including age and gender, were added to produce a final predictive score.

Key findings of the study

“We wanted to predict what would happen in the next six years, and we found that AI could confidently and accurately predict what would happen to people’s cognition six years before they showed any symptoms,” says computer scientist Ioannis Paschalidis of Boston University, who co-authored the study.

“Since there is currently no cure for Alzheimer’s, early detection is important, as it helps manage and reduce symptoms, opening the door to developing a fully effective treatment.”