AI Model Predicts Alzheimer's Onset with Speech Analysis

A new AI-based model developed by Boston University predicts Alzheimer's disease onset by analyzing speech patterns. Trained on audio recordings, it achieved 78.5% accuracy in forecasting disease progression. Researchers highlight its potential for early diagnosis via an accessible app, crucial for preventing significant cognitive decline.

A groundbreaking AI-based model developed by researchers at Boston University has shown promise in predicting the onset of Alzheimer's disease through speech analysis, the developers announced.

The model, trained on audio recordings of patients with mild cognitive impairment, achieved an impressive 78.5 per cent accuracy in forecasting whether individuals would remain stable or progress to dementia within a six-year timeframe.

Alzheimer's disease, the most pervasive form of dementia, severely impacts one's daily activities by deteriorating memory and thinking abilities.

The research team utilized initial interviews with 166 patients aged 63-97, applying machine learning techniques to identify crucial patterns in speech, demographics, and diagnosis progression.

The findings, published in the journal Alzheimer's and Dementia, revealed the model analyses both the content of speech and sentence structure rather than focusing solely on enunciation or speed.

Lead researcher Ioannis C. Paschalidis explained, "We combine information from the audio recordings with basic demographics - age, gender, etc., to produce a final score indicating the likelihood of dementia progression."

Despite challenges such as low-quality recordings, the model demonstrated significant predictive capabilities, the researchers noted.

The importance of early detection is emphasized, as current diagnostic methods often identify Alzheimer's only after considerable cognitive decline is noticeable.

The team aims to develop an app to make this predictive model accessible to patients in remote regions, enhancing early screening opportunities.

(This story has not been edited by Devdiscourse staff and is auto-generated from a syndicated feed.)