A new artificial intelligence (AI) model developed by Boston University researchers shows promising potential in predicting whether individuals with mild cognitive impairment will develop Alzheimer's disease. By analyzing speech content, the model achieved an impressive accuracy rate of 78.5%. This breakthrough could lead to earlier diagnoses and more accessible cognitive impairment screening without the need for expensive tests.

Traditionally, diagnosing Alzheimer's disease involves a series of assessments, including interviews, brain imaging, and blood tests. However, by the time these tests are conducted, the disease may have already progressed significantly. The new AI model offers a non-invasive method to monitor dementia risk, enabling healthcare professionals to intervene earlier and potentially slow the disease's progression.

Powered by machine learning, a subset of AI that allows programs to independently analyze data, the model evaluates speech content, rather than just acoustic features. The research team trained the model using data from the long-running Framingham Heart Study, which provided extensive information on cognitive decline. The model was trained on audio recordings and demographic data from 166 individuals diagnosed with mild cognitive impairment. After training, it accurately predicted whether participants would remain stable or transition to dementia.

"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 research team hopes that this technology will not only improve early diagnosis but also make cognitive impairment screening more accessible by automating parts of the process.

By focusing on the content of the speech, rather than acoustic features, the model demonstrates its ability to make accurate predictions even when dealing with low-quality recordings and background noise. This technology has the potential to provide efficient and automated dementia diagnosis, reducing the need for human involvement and increasing screening capacity. It could bring care to patients in remote areas and improve access to treatment and care for the majority who never receive a formal dementia diagnosis.

Rhoda Au, a coauthor on the study, emphasized the potential of AI to create "equal opportunity science and healthcare." The researchers plan to expand their study beyond speech analysis to include patient drawings and data on daily life patterns to further improve the model's predictive accuracy. They also aim to explore the use of natural conversations and smartphone apps to diagnose dementia in future research.

Funding for this research was provided in part by the National Science Foundation, the National Institutes of Health, and the BU Rajen Kilachand Fund for Integrated Life Science and Engineering. The findings have been published in Alzheimer's & Dementia, the journal of the Alzheimer's Association.

This innovative AI model offers hope for earlier diagnosis and improved accessibility in screening for Alzheimer's disease. With further advancements, it has the potential to revolutionize dementia care and provide valuable insights into predicting and managing this devastating condition.