An artificial intelligence (AI) model analyzing transcripts of speech from past cognitive tests predicted the progression of cognitive impairment to Alzheimer’s disease within six years with more than 78% accuracy. The NIA-funded study results were published in Alzheimer’s and Dementia.
Scientists applied an AI speech analysis system to evaluate samples from transcripts of cognitive tests given to 166 participants in the Framingham Heart Study (FHS), one of the nation’s longest-running, long-term research studies on heart health. The 166 participants all had been diagnosed with mild cognitive impairment (MCI). The speech transcripts were divided into two subgroups based on level of MCI progression, with 90 progressive cases and 76 stable ones.
This study was part of a series conducted by a Boston University-led research team seeking to fine-tune and expand usage of their speech analysis AI as a potential tool to automate elements of dementia diagnosis. The AI system analyzed just the language structure of automated transcripts of voice recordings of participants’ speech during their cognitive tests; it did not evaluate any acoustic or vocal properties.
The results showed the AI model successfully predicted, with an accuracy of 78.2%, which participants progressed from MCI to Alzheimer’s within six years. The researchers view this as validating the potential of AI speech analysis as a convenient, inexpensive cognitive testing resource that can be used remotely to complement other tests and biomarkers. They suggest their method is more accurate than other noninvasive tests and can help medical professionals and clinical trial managers better identify people at risk for MCI progressing to Alzheimer’s.
A limitation of this study is that FHS participants were predominantly White; thus, the team hopes to expand this research into larger, more diverse study populations in future studies.
This research was supported in part by NIA grants AG008122, AG016495, AG062109, AG068753, and AG072654.
These activities relate to NIH’s AD+ADRD Research Implementation Milestone 9.H,“Biomarkers: Early detection measures.”
Reference: Amini S, et al. Prediction of Alzheimer’s disease progression within 6 years using speech: A novel approach leveraging language models. Alzheimer’s & Dementia. 2024;20(8):5262-5270. doi: 10.1002/alz.13886.
