By Clinical Research News Staff 

January 20, 2026 | Artificial intelligence (AI) is reshaping how clinical researchers study sleep, which has been tested with either simple self-reporting sleep logs or cumbersome sleep studies that take place in a lab or clinic. A new AI-driven sleep-staging framework called BIDSleep was developed at the University of Massachusetts Amherst. This new system illustrates how consumer wearables may begin to close that methodological gap and expand the scale of sleep research beyond the clinic. 

BIDSleep repurposes the Apple Watch into a research-grade sleep-staging device capable of distinguishing light, deep, and rapid eye movement (REM) sleep. Led by Joyita Dutta, Ph.D., a professor of biomedical engineering at University of Massachusetts Amherst, the work targets a persistent challenge in clinical research: how to collect reliable, longitudinal sleep data in real-world settings without relying on overnight polysomnography. 

Polysomnography remains the gold standard for sleep studies, but its cost and complexity make it impractical for large studies or repeated measurements. On the other hand, sleep diaries and questionnaires are inexpensive and scalable but highly variable at the individual level. BIDSleep is designed to make the process easier and more accessible while maintaining accuracy. It extracts clinically relevant information from heart rate and accelerometry signals already collected by wrist-worn devices. 

The framework was evaluated in a study of 47 healthy adults, each monitored for up to seven consecutive nights using an Apple Watch Series 6 alongside a Dreem 2 EEG headband as a reference comparator. Results, published in IEEE Transactions on Biomedical Engineering (DOI: 10.1109/TBME.2025.3612158), showed the AI correctly identified sleep stage 71% of the time—outperforming several established time-series and heart rate–based approaches commonly used in sleep research. 

The results extend beyond accuracy. The model demonstrated particular strength in identifying deep sleep, a stage closely linked to aging, cognitive decline, and neurodegenerative disease. It also captured metrics such as sleep efficiency and sleep onset latency, measures that are often unreliable when derived from self-report but are frequently used as endpoints in observational studies and trials. 

Dutta’s broader research program focuses on the relationship between sleep disruption and Alzheimer’s disease, integrating wearable-derived sleep metrics with neuroimaging, blood biomarkers, and genetic risk profiling. The long-term aim is to understand how early changes in sleep architecture intersect with amyloid and tau pathology during preclinical stages of dementia—precisely the window when disease-modifying therapies are thought to be most effective. The team also sees potential applications in studies for mood disorders and evaluating the effects of medical procedures or therapies.  

To read the full story written by Deborah Borfitz, visit Diagnostics World News.