AI Sleep-Staging Framework Achieves 71% Accuracy Using Apple Watch Data
Researchers developed BIDSleep, an AI framework that uses Apple Watch data to accurately classify sleep stages, achieving 71% accuracy in a clinical study. Separately, startup SOND launched AI-powered earbuds that track physiological signals to enhance sleep. Both advances demonstrate the expanding role of AI and wearables in sleep science.
Researchers have developed an AI-driven framework called BIDSleep that enables accurate sleep-stage classification using data from a consumer smartwatch. The system was found to correctly identify sleep stages 71% of the time, outperforming several established methods and offering a scalable alternative to lab-based sleep studies.
Developed at the University of Massachusetts Amherst, BIDSleep repurposes the Apple Watch into a research-grade device capable of distinguishing between light, deep, and rapid eye movement (REM) sleep. The framework analyzes heart rate and accelerometry signals already collected by the wrist-worn device. In a study of 47 healthy adults monitored for up to seven consecutive nights, the system’s accuracy was validated against a Dreem 2 EEG headband, a reference comparator. The results were published in IEEE Transactions on Biomedical Engineering.
The model showed 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 often unreliable when derived from self-report but frequently used as study endpoints. The research team’s broader program focuses on the relationship between sleep disruption and Alzheimer’s disease, integrating wearable sleep metrics with neuroimaging, blood biomarkers, and genetic risk profiling. Potential applications include studies on mood disorders and evaluating the effects of medical procedures or therapies.
In the consumer space, a startup called SOND has introduced Dreambuds, described as a closed-loop, in-ear system that tracks a dozen physiological signals and responds in real time. Founded by former Bose executives, the company positions the device as an active sleep enhancement platform equipped with precision sensors, generative audio, and a personal AI sleep coach. The earbuds aim to help users fall asleep faster, stay asleep longer, and wake up feeling restored.
These developments highlight a growing trend toward using artificial intelligence and wearable sensors to make sleep monitoring more accessible and precise. While clinical validation remains critical for research applications, consumer devices are beginning to incorporate advanced biosignal tracking, moving beyond simple noise-masking earbuds or basic sleep logs.