Researchers at Boston University have created a groundbreaking artificial intelligence (AI) model capable of predicting whether a person with mild cognitive impairment will develop Alzheimer’s disease within six years, based solely on their speech. This model has shown an impressive accuracy rate of 78.5%, offering a non-invasive, accessible method for early diagnosis. The findings have been published in the journal Alzheimer’s & Dementia.
Alzheimer’s disease is the most common cause of dementia and typically progresses through a long phase where only subtle cognitive changes are apparent. People with mild cognitive impairment are at a higher risk of developing Alzheimer’s, with annual conversion rates ranging from 3% to 15%.
Early and accurate prediction of who will transition from mild cognitive impairment to Alzheimer’s is vital for timely treatment and inclusion in clinical trials for new drugs. Traditional methods for diagnosing Alzheimer’s, such as brain imaging and cerebrospinal fluid tests, are invasive, expensive, and not easily accessible in all regions. In contrast, analyzing speech during neuropsychological tests is a less invasive and potentially more scalable approach.
The Boston University researchers conducted their study using data from the Framingham Heart Study, which has been recording neuropsychological test interviews since 2005. The cohort for this study consisted of 166 individuals with cognitive complaints, including 59 men and 107 women, with a median age of 81 years. These participants underwent neuropsychological tests designed to assess various cognitive domains such as memory, language, visuospatial skills, abstract reasoning, and attention. Each test session, lasting about an hour, was audio-recorded and stored in the .wav format.
The researchers transcribed the audio recordings into text using automated speech recognition software. Each utterance was then diarized, meaning it was attributed to either the participant or the examiner, and categorized into specific subtests like the Boston Naming Test or the Wechsler Memory Scale. The text data was processed using the Universal Sentence Encoder, a deep learning model that transforms text into numerical vectors representing semantic content.
To predict whether individuals with mild cognitive impairment would progress to Alzheimer’s disease within six years, the researchers employed logistic regression models. They generated embedding vectors from the transcribed text and trained the models on these vectors, along with demographic information such as age, sex, and education level. The model’s performance was evaluated using stratified group k-fold cross-validation, ensuring that the data was split into multiple folds for training and testing to validate the results comprehensively.
The AI model developed by the researchers achieved an accuracy rate of 78.5% and a sensitivity of 81.1% in predicting the progression from mild cognitive impairment to Alzheimer’s disease within six years. Sensitivity refers to the model’s ability to correctly identify individuals who will progress to Alzheimer’s, while specificity, at 75%, measures its accuracy in identifying those who will not progress. These results indicate a strong predictive power, particularly in identifying future Alzheimer’s patients.
The analysis revealed that speech features extracted from the neuropsychological test recordings were robust predictors of disease progression. The inclusion of text features alone outperformed traditional neuropsychological test scores and demographic factors. This suggests that subtle changes in speech patterns and language use can provide valuable insights into cognitive decline, even before more obvious symptoms appear.
“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 (Yannis) Paschalidis, the director of the Boston University Rafik B. Hariri Institute for Computing and Computational Science & Engineering. “It shows the power of AI.”
Interestingly, the study also found that older women, individuals with lower education levels, and those carrying specific genetic markers, such as the apolipoprotein E gene allele, were more likely to progress to Alzheimer’s. These findings align with existing research on the risk factors for Alzheimer’s disease, reinforcing the validity of the AI model’s predictions.
While the AI model shows promise, there are limitations to consider. The study’s cohort was predominantly White, limiting the generalizability of the findings to more diverse populations. The specificity of the model, although reasonable, still leaves room for improvement to reduce the cost of clinical trials by better identifying candidates for new treatments.
Furthermore, the model’s reliance on speech data means that variations in dialect, language proficiency, and cultural differences could impact its accuracy. Future research should aim to validate these findings across more diverse and larger populations and explore the inclusion of other types of data, such as patient drawings and daily life patterns, to enhance predictive accuracy.
As the study’s co-author, Rhoda Au, notes, AI has the potential to create “equal opportunity science and healthcare,” overcoming biases and resource limitations. This technology can democratize access to early diagnosis and treatment, making it available to a broader population.
“Technology can overcome the bias of work that can only be done by those with resources, or care that has relied on specialized expertise that is not available to everyone,” said Au, a professor of anatomy and neurobiology. One of the most exciting findings was “that a method for cognitive assessment that has the potential to be maximally inclusive—possibly independent of age, sex/gender, education, language, culture, income, geography—could serve as a potential screening tool for detecting and monitoring symptoms related to Alzheimer’s disease.”
The researchers plan to expand their study to include data from more natural conversations, rather than just structured neuropsychological tests, potentially through a smartphone app. This approach could make the AI model even more accessible and practical for widespread use. Additionally, they aim to incorporate other types of data to further improve the model’s accuracy.
“Digital is the new blood,” says Au. “You can collect it, analyze it for what is known today, store it, and reanalyze it for whatever new emerges tomorrow.”
“We hope, as everyone does, that there will be more and more Alzheimer’s treatments made available,” Paschalidis added. “If you can predict what will happen, you have more of an opportunity and time window to intervene with drugs, and at least try to maintain the stability of the condition and prevent the transition to more severe forms of dementia.”
The study, “Prediction of Alzheimer’s disease progression within 6 years using speech: A novel approach leveraging language models,” was authored by Samad Amini, Boran Hao, Jingmei Yang, Cody Karjadi, Vijaya B. Kolachalama, Rhoda Au, and Ioannis C. Paschalidis.