Osegi
AI·2 min read

AnySleep: Open-Source AI Analyzes Sleep Second by Second

Researchers at FH Aachen have built AnySleep, an open-source AI model that reads brain activity in intervals shorter than the standard 30 seconds to analyze sleep more precisely.

Illustration of an AI model analyzing EEG brain-wave data to classify sleep stages in short intervals.
Image: t3n Magazin
SUMODialoediviewergiftdetect

Researchers at FH Aachen University of Applied Sciences have developed AnySleep, an open-source AI model that analyzes human sleep far more finely than conventional methods. By detecting even brief shifts in brain activity, it can map sleep in much shorter intervals. The results were recently published in npj Digital Medicine, part of the Nature Portfolio.

Beyond the 30-second window

Sleep studies usually record brain activity with electroencephalography (EEG). Specialists then split a night's data into 30-second segments and assign a sleep stage to each one, producing a step-like picture of how sleep progresses. According to the team, AnySleep can resolve shorter intervals and therefore pick up very brief changes in brain activity, producing a detailed overall picture of the night's sleep phases.

"Short wake events, which often last only a few seconds, can provide important clues in sleep research. If the evaluation interval is too long, these changes are not sufficiently taken into account and the result becomes imprecise," says study author Stephan Bialonski. To train the AI, the researchers used sleep data from roughly 20,000 participants whose age, living circumstances and health status varied widely.

Potential for medicine, with limits

The finer view has already revealed differences in sleep structure between healthy people and people with illnesses. Bialonski concedes, however, that the method is not yet suitable for diagnosing diseases. "But AnySleep already provides indications of certain conditions and helps us learn more about sleep," he says. The project is to continue in cooperation with several clinics.

The source code is publicly available and can in theory be installed on any computer. FH Aachen also offers SomnoBot, a browser interface that lets people use the tool without coding skills. According to the press release, sensitive data never leaves the local computer, which makes the tool suitable for strict clinical research settings from a data protection standpoint.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:anysleepfh aachensleep analysisopen sourceeegnpj digital medicine

More from this day

See all →

More in: AI

See all →