Search Here

Researchers at Boston University have unveiled an artificial intelligence program that uses speech analysis to predict the risk of Alzheimer's disease in individuals with mild cognitive impairment.

The program boasts an accuracy rate of 78.5 percent in predicting whether patients will experience stable cognitive function or progress to dementia associated with Alzheimer's over the next six years.

The AI model, detailed in Alzheimer's & Dementia, a journal of the Alzheimer's Association, marks a significant advancement in predictive healthcare technology. By leveraging machine learning techniques, the program analyzes audio recordings from initial interviews with 166 participants aged 63 to 97, drawn from the renowned Framingham Heart Study. Among these participants, 76 remained cognitively stable, while 90 showed progressive cognitive decline.

"We wanted to predict what would happen in the next six years-and we found we can reasonably make that prediction with relatively good confidence and accuracy," said Ioannis Paschalidis, Director of the BU Rafik B. Hariri Institute for Computing and Computational Science & Engineering.

The model integrates speech content with basic demographic data such as age and gender, eschewing acoustic features like pronunciation or speed. Despite challenges such as low-quality recordings and background noise, the AI successfully discerned predictive patterns solely from conversational content, Neuroscience News reported.

Rhoda Au, a coauthor and professor at BU's Chobanian & Avedisian School of Medicine, emphasized the potential for AI-driven assessments to democratize healthcare access.

"Technology can overcome the bias of work that can only be done by those with resources," Au said, noting AI's role in expanding diagnostic capabilities beyond traditional clinical settings.

The researchers envision future applications where AI-powered assessments could be conducted remotely through smartphone apps, offering widespread screening and monitoring for Alzheimer's-related symptoms. This approach could bridge gaps in dementia diagnosis globally, where many affected individuals currently lack access to formal healthcare.

Looking ahead, Paschalidis aims to enhance the model's accuracy by incorporating additional data sources beyond speech, such as patient drawings and daily activity patterns recorded in the Framingham tests. He anticipates further advancements in AI technology will continue to refine predictive capabilities, potentially transforming dementia care and treatment strategies.