Scientists Develop Speech Recognition Tool To Predict Alzheimer's Onset

A new AI-based model could predict the onset of Alzheimer's disease by analysing an individual's speech, the developers said.

Trained on audio recordings of patients with mild cognitive impairment -- early stages of memory loss, the model achieved 78.5 per cent accuracy in forecasting whether patients would remain stable or progress to dementia within six years, according to the researchers.

Alzheimer's disease is the most common form of dementia and impacts one's everyday activities by impairing memory and thinking.

The researchers at Boston University, US, used recordings of initial interviews of 166 patients aged 63-97 and trained the model using machine learning to discern patterns between speech, demographics, diagnosis, and how their condition was worsening.

The model analyses interview content such as spoken words and sentence structure, rather than speech features such as enunciation or speed, showed the study published in the journal Alzheimer's and Dementia.

"We combine the information we extract from the audio recordings with some very basic demographics - age, gender, and so on - and we get the final score," said Ioannis C. Paschalidis, a professor of engineering and the study's corresponding author.

"You can think of the score as the likelihood, the probability, that someone will remain stable or transition to dementia. It had significant predictive ability," said Paschalidis.

The researchers said that the model was able to perform well, despite challenges like low-quality recordings and background noise.

The researchers emphasised that early prediction is crucial as current diagnostic tests often identify Alzheimer's disease only after significant cognitive decline has occurred such as memories starting to slip away and personality traits beginning to shift.

The team aims to make their model accessible through an app to make it accessible for patients in remote areas, potentially increasing the number of people getting screened.
 

(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated feed.)