Overview: A new machine learning algorithm can accurately detect cognitive impairment by analyzing voice recordings.
Source: Boston University
It takes a lot of time — and money — to diagnose Alzheimer’s disease. After lengthy personal neuropsychological examinations, clinicians must transcribe, assess and analyze each response in detail.
But researchers at Boston University have developed a new tool that can automate the process and eventually let it go online. Their machine learning-based computer model can detect cognitive impairment using audio recordings of neuropsychological tests — no need for a face-to-face appointment.
Their findings were published in Alzheimer’s and Dementia: The Journal of the Alzheimer’s Association†
“This approach brings us one step closer to early intervention,” said Ioannis Paschalidis, a co-author of the paper and a BU College of Engineering Distinguished Professor of Engineering.
He says faster and earlier detection of Alzheimer’s could lead to larger clinical trials targeting individuals in early stages of the disease and potentially enabling clinical interventions that slow cognitive decline: “It could form the basis of an online tool that can reach anyone and could increase the number of people getting screened early.”
The research team trained their model using audio recordings of neuropsychological interviews of more than 1,000 individuals in the Framingham Heart Study, a long-term BU-led project looking at cardiovascular disease and other physiological disorders.
Using automated online speech recognition tools (think “Hey, Google!”) and a machine learning technique called natural language processing that helps computers understand text, they had their program transcribe the interviews and then encode them into numbers.
A definitive model was trained to assess the likelihood and severity of a person’s cognitive impairment using demographics, the text encodings, and real diagnoses from neurologists and neuropsychologists.
Paschalidis says the model was not only able to accurately distinguish between healthy individuals and people with dementia, but also detect differences between people with mild cognitive impairment and dementia. And it turned out that the quality of the recordings and the way people spoke—whether their speech passed quickly or stuttered constantly—were less important than the content of what they said.
“We were surprised that speech flow or other audio features are not so important; you can transcribe interviews quite well automatically and rely on AI text analysis to assess cognitive impairment,” said Paschalidis, who is also the new director of BU’s Rafik B. Hariri Institute for Computing and Computational Science & Engineering.
While the team has yet to validate the results with other data sources, the findings suggest that their tool could help clinicians diagnose cognitive impairment using audio recordings, including those from virtual or telehealth appointments.
Screening before the onset of symptoms
The model also provides insight into which areas of neuropsychological testing may be more important than others in determining whether a person has impaired cognition. The researchers’ model divides the exam transcripts into different sections based on the clinical tests performed.
For example, they found that the Boston Naming Test — in which clinicians ask individuals to label a picture with one word — is the most informative for an accurate diagnosis of dementia.
“This allows clinicians to allocate resources in a way that allows them to do more screening even before symptoms start,” says Paschalidis.
Early diagnosis of dementia is not only important for patients and their caregivers to create an effective treatment and support plan, but it is also crucial for researchers working on therapies to slow and prevent the progression of Alzheimer’s disease. to prevent.
“Our models can help clinicians assess patients in terms of their chances of cognitive decline,” says Paschalidis, “and then tailor resources best to them by conducting further testing on those at greater risk of dementia.”
Would you like to participate in the research effort?
The research team is looking for volunteers to complete an online survey and submit an anonymous cognitive test. The results will be used to provide personalized cognitive assessments and will also help the team refine their AI model.
About this research news on AI and Alzheimer’s disease
Author: Molly Gluck
Source: Boston University
Contact: Molly Gluck – Boston University
Image: The image is in the public domain
Original research: Closed access.
“Automated Detection of Mild Cognitive Impairment and Dementia from Voice Recordings: An Approach to Natural Language Processing” by Ioannis Paschalidis et al. Alzheimer & Dementia
Automated Detection of Mild Cognitive Impairment and Dementia from Speech Recordings: A Natural Language Processing Approach
Automated computational assessment of neuropsychological tests would enable widespread, cost-effective screening for dementia.
A novel natural language processing approach has been developed and validated to identify different stages of dementia based on computerized transcription of digital voice recordings of subjects’ neuropsychological tests conducted by the Framingham Heart Study (n = 1084). Transcribed sentences from the test were coded into quantitative data and various models were trained and tested using this data and the demographics of the participants.
The mean area under the curve (AUC) on the retained test data reached 92.6%, 88.0%, and 74.4%, respectively, for distinguishing normal cognition from dementia, normal or mild cognitive impairment (MCI) from dementia, and normal from MCI.
The proposed approach provides a fully automated identification of MCI and dementia based on a registered neuropsychological test, providing the opportunity to develop a remote screening tool that can be easily adapted to any language.