GabyResearchers at Boston University say we still don’t know exactly what causes Alzheimer’s disease. But we do know its consequences, and they’re getting better at spotting early — even hidden — signs in human speech.
The university team developed a new algorithm based on artificial intelligence. It analyzes the language patterns of people with mild cognitive impairment (MCI) and can predict whether their condition will progress to Alzheimer’s disease within six years with 78.5 percent accuracy.
This work continues a previous study in which the team trained the model to detect cognitive impairments using voice recordings from over 1,000 people.
What the research showed
The new algorithm was trained on transcribed audio recordings from 166 people with MCI, aged 63 to 97, the publication Science Alert reported. Because the researchers already knew which participants had gone on to develop Alzheimer’s disease, they could use the machine-learning model to search transcribed speech for features that linked to the 90 individuals whose cognition declined to the point of an Alzheimer’s diagnosis.
After training, they applied the algorithm in reverse: using previously unseen transcribed language samples to predict Alzheimer’s risk. They added other factors, such as age and gender, for the final assessment.
“One can consider the assessment as the probability that a person’s condition will remain stable or that they will develop dementia,” said the study’s lead author, informatics specialist Ioannis Paschalidis.
“We wanted to predict what would happen over the next six years, and found we could make that forecast with fairly good confidence and accuracy. It shows the power of AI,” he added.
Why is this important?
Alzheimer’s disease is currently incurable, which raises a fair question: what’s the point of early detection if the outcome doesn’t change? The researchers say some treatments can help manage symptoms, and diagnosing earlier means management can start sooner.
Early detection also gives scientists more time to study disease progression and develop treatments, and lets at-risk people enroll in clinical trials earlier.
The team says AI detection will keep improving. This kind of test could be done quickly and cheaply at home—no special equipment, no injections or biological samples—just a voice recording. In the future it could come as a smartphone app.
“If you can anticipate what will happen, you have more time to try medications or other interventions; at minimum you can try to maintain stability and prevent progression to more severe dementia,” Paschalidis said.
The method could also help researchers understand why mild cognitive impairment progresses to Alzheimer’s in some people but not others.
The results of the study were published in the journal Alzheimer’s & Dementia.
