Uncovering the Hidden Signs of Alzheimer's with AI-Powered Imaging
Joyita Dutta, professor of biomedical engineering, leads the Biomedical Imaging and Data Science Lab, which applies machine learning and data science to problems in neuroimaging and brain health. Her lab focuses on using non-invasive, real-world data to improve early diagnosis and monitoring of neurodegenerative disorders such as Alzheimer’s and frontotemporal dementia.
Sleep disruption is a well-established hallmark of Alzheimer’s disease, often appearing years before memory loss and other cognitive symptoms.
Supported by a National Institutes of Health (NIH) grant and a Massachusetts AI and Technology Center for Connected Care in Alzheimer’s Disease (MassAITC) pilot award, Dutta is investigating whether consumer wearable devices can help detect subtle sleep changes that correspond with early biological markers of Alzheimer’s. While the current gold standard for sleep studies is through lab-based assessments, they are also complex, expensive, and require manual data analysis by a specialist.
Furthermore, because of their cost and complexity, the duration of most sleep studies is only one night, meaning researchers don’t have the benefit of analyzing data from multiple sessions over time.
On the other hand, as Dutta notes: “Many people already wear smartwatches to sleep these days. Imagine receiving an alert from your smartwatch advising you to see a neurologist. That could be the direction we are headed.”
In a research study Dutta is leading, participants wear multiple consumer sleep trackers—including the Apple Watch, Oura Ring, and CGX Patch—for a week rather than undergoing a single-night lab sleep study. Dutta and her team then compare wearable-derived sleep metrics with blood levels of amyloid and tau proteins, biomarkers closely linked to the early stages of Alzheimer’s. Assessments will be repeated after two years to identify trends that may signal cognitive decline.
The project aims to address a major public health gap: determining who should be screened for Alzheimer’s and when. Wearables could offer a non-invasive, relatively low-cost way to monitor people over long periods and flag those who may benefit from clinical testing.
As part of this work, Dutta and her group recently developed an app that turns consumer Apple Watches into sophisticated sleep-staging tools. The app, called BIDSleep, collects high-resolution instantaneous heart- rate data—which varies across sleep stages—and feeds it into an AI model trained to identify light, deep, and REM sleep. The researchers report that their model correctly identified sleep stages about 71% of the time, outperforming several established approaches and doing particularly well at detecting deep sleep—an important metric because deep sleep tends to decline with age.
“Our goal was to get as rugged as possible with a non-specialized consumer wearable device, which is the Apple Watch,” says Dutta.
Dutta also notes that, for her ongoing Alzheimer’s research, the existing monitoring technology cannot capture sleep data from naps, which are often unplanned. In contrast, the broad availability and round-the-clock wearability of smartwatches make them particularly well-suited for studying all forms of sleep.
Tzu-An Song, a postdoctoral fellow in Dutta’s lab and first author on the sleep-staging paper, notes the importance of clinically relevant measures beyond raw classification accuracy: “Overall accuracy matters, but sometimes we also need to look at the clinical metrics like sleep efficiency and sleep onset latency, total sleep time.” He adds that, on those measures, the BIDSleep approach performed strongly: “Our method works better for basically all of these metrics,” he says.
The BIDSleep software and its AI code are made available for other researchers, with the intent of simplifying data export and analysis. Dutta says, “Ultimately, we’d love for researchers and clinicians to use this app, which is why we created it in a style where you can easily port the data and get multi-night information out of it.”
While blood-based Alzheimer’s assays are improving, broad screening remains challenging. Publicly accessible consumer wearables, combined with models like those from Dutta’s lab, could serve as a critical early-warning layer, identifying at-risk individuals whose sleep patterns warrant further evaluation.