Summary: Researchers are developing a machine learning model aimed at early detection of Alzheimer’s dementia. This model, accessible via smartphones, can distinguish between Alzheimer’s patients and healthy individuals with 70-75% accuracy.
By focusing on speech patterns rather than content, the tool can provide invaluable early indicators, which can initiate early treatment and slow disease progression.
While not a substitute for healthcare professionals, telehealth can enhance access and help overcome geographic or language barriers.
- A machine learning model can distinguish Alzheimer’s patients from healthy individuals with 70-75% accuracy.
- The tool examines acoustic and linguistic speech characteristics rather than specific words to diagnose the disease.
- An application of this model can be a simple and accessible screening tool on smartphones, providing early indicators of Alzheimer’s.
Source: University of Alberta
Researchers are already trying to transform Alzheimer’s dementia with a machine learning (ML) model into a simple diagnostic tool that anyone with a smartphone could one day use.
The model was able to distinguish Alzheimer’s patients from healthy controls with 70 to 75 percent accuracy, which is promising for the more than 747,000 Canadians with Alzheimer’s or other dementias.
Alzheimer’s dementia can be challenging to recognize in its early stages, as symptoms often begin very subtle and can be confused with age-related memory issues. But the researchers said that the earlier possible cases are detected, the sooner patients begin to take action.
“Before, you needed lab work and medical imaging to detect brain changes. This takes time, it’s expensive, and no one has explored this before,” said Professor Eleni Strolia of the Department of Computing Science, who was involved in creating the model.
“If you can use mobile phones to find a previous referral, that means communicating the patient’s relationship with their doctor. It can start treatment earlier, and we can start with simple interventions at home and even mobile devices to slow its progression.
A screening tool does not replace healthcare professionals. But in addition to helping with early detection, telehealth provides a convenient way to identify potential risks for patients who face geographic or language barriers to access services in their area, said Zehra Shah, a master’s student in the Department of Computing. Science and the first author of the paper.
“We can think about identifying patients using this type of technology based on speech alone,” says Shah.
While the research team has previously looked at the language used by Alzheimer’s dementia patients, for this project they examined language-agnostic acoustic and speech characteristics of the language rather than specific words.
“The first task involves listening to what the person is saying, understanding what he is saying and understanding the meaning. This is a simple computational problem to solve,” Strolia said. “Now listen to the voice. There are certain properties in the way people speak that transcend language.
“It’s more powerful than any version of the problem we’ve been solving before,” Strolia added.
The researchers started with the speech characteristics that the doctors had extracted from the patients with Alzheimer’s dementia. These patients tended to speak slowly, with frequent pauses or interruptions in their speech.
They typically use shorter words, and often have reduced intelligibility in their speech. Researchers have found ways to translate these features into speech features that the model can check.
While the researchers focused on English and Greek speakers, “this technology has the potential to be used across languages,” Shah said.
And while the model itself is complex, the end-user experience for the device that incorporates it couldn’t be simpler.
“A person talks to the device, it analyzes it and it makes a prediction: either yes, the person has Alzheimer’s or no, they don’t,” said Russ Greiner, the paper’s coauthor and a professor in the Department of Computing Science. . That information can then be brought to the health care professional to determine the best course of action for the individual.
Both Greiner and Stroulia lead a computational psychiatry research group at the U of A, whose members have developed similar AI models and tools to diagnose mental illnesses such as PTSD, schizophrenia, depression and bipolar disorder.
“Anything we can do to improve clinical processes, inform treatments and manage diseases at a lower cost is great,” says Stroulia.
So machine learning and Alzheimer’s disease research news
Author: Adriana McPherson
Source: University of Alberta
Contact: Adriana McPherson – University of Alberta
Image: Image credited to Neuroscience News.
Preliminary study: The findings will be presented at ICASSP 2023—2023 IEEE International Conference on Acoustics, Speech and Signal Processing.