Early detection remains a key objective for Alzheimer’s disease researchers. Developing accurate screening tools that are easy to administer play a key role in finding the disease as early as possible to allow for interventions. A study recently published in Alzheimer’s and Dementia shows encouraging findings related to an AI model that was used to analyze transcripts of voice recordings.

What the results showed

The Framingham Heart Study began in 1948 with an emphasis on stroke and heart disease. Over time, the research evolved to include observing the aging of the brain and tracking those experiencing cognitive changes.

“We have continued this study to include new cohorts, including younger generations of the original participants, to look for risk factors at an earlier age,” says Rhoda Au, Ph.D., Investigator – Framingham Heart Study, Departments of Anatomy & Neurobiology, Neurology, Medicine and Epidemiology, Boston University Chobanian & Avedisian School of Medicine and the School of Public Health. “Starting in 2005, we began digitally recording spoken responses to cognitive tests. Around 2008, we realized these recordings were not just useful for scoring purposes on the tests they were administered for but also were data themselves.”

For this study, the voice recordings of 166 participants who had all been diagnosed with mild cognitive impairment (MCI) became transcripts which researchers then applied to an AI speech analysis system. The AI model achieved 78.2% accuracy when identifying which participants progressed from MCI to Alzheimer’s disease within six years.

“These results help us build the case that digital voice alone is showing validation when it comes to predicting Alzheimer’s disease,” explains Dr. Au. “This is important because it offers the potential of a convenient, inexpensive testing resource that could be used as a screening tool.”

The importance of digital data

Cognition is something that fluctuates throughout the day. Because it isn’t static, there’s a lot of variability when it comes to accurately identifying if there is an issue.

“When you talk to a patient, they may give examples of memory concerns such as getting lost going to a familiar place or forgetting to turn off the stove,” shares Dr. Au. “The problem is these issues likely don’t happen all of the time. If you talk to family members, they will also be able to give examples of the patient’s memory problems that they noticed. Often these family responses vary from each other and from what the patient is sharing. This can make it difficult to identify a single pattern but taken together result in a more accurate report of an actual problem.”

Digital clinical assessments, such as those using digital voice, can take the guesswork out of identifying patterns of cognitive change.

“We’ve made significant investments in imaging and blood-based biomarkers, which has allowed us to detect the tau and amyloid proteins that are the pathology of Alzheimer’s disease,” says Dr. Au. “But we need to make those same investments of time, energy and money to tap into the tremendous potential of digital to detect clinical symptoms.”

While using digital information is more complex than traditional tests for a disease, such as a blood test, it offers the potential for much earlier detection of the disease compared to long-used testing methods.

What’s ahead for this technology

“We’re in the process of developing a white-labeled application to record digital voice on a smartphone with the goal of allowing anyone, anywhere to collect data for research,” explains Dr. Au. “This will ultimately allow us to work across the population to get enough information to achieve an accurate algorithm for the AI.”

With how pervasive smartphones are across the world, having a validated method to analyze digital voice would create a novel way to screen for Alzheimer’s disease on a global scale.

“If we have a reliable, highly correlated indicator for Alzheimer’s disease, we’ll have the ability to detect cognitive changes sooner, providing the opportunity for earlier interventions to slow progression or delay onset,” adds Dr. Au.