PROJECT TEAM
Poul Jørgen Jennum, Professor, MD, Danish Center for Sleep Medicine, Rigshospitalet
Mathias Perslev PhD student, Department of Computer Science, University of Copenhagen
Christian Igel, Professor, Dr Habil, Department of Computer Science, University of Copenhagen
Miki Nikolic PhD, engineer, Danish Center for Sleep Medicine, Rigshospitalet
Sune Darkner Associate Professor, Department of Computer Science, University of Copenhagen
THE NEED
More than 15% of Danes suffer from sleep disorders such as sleep apnea, which are linked to severe health risks. Diagnosis requires sleep recordings (PSGs), but the manual evaluation process is slow, labor-intensive, and limits how many patients can be diagnosed and treated. Current automated tools lack the accuracy and robustness needed for clinical use.
THE SOLUTION
U-Sleep is a clinically robust deep learning model for automatic sleep staging. It replaces 2 –3 hours of manual work per PSG with fast, reliable annotations—freeing up staff time and increasing diagnostic throughput. Trained on a large, diverse dataset, U-Sleep is accurate across patient types, hardware, and clinical setups.
U-Sleep: Automatic Clinical Sleep Staging
Call 3 - 2022
500.000 DKK
Clinical Area
Sleep medicine, neurology
Technology
AI-based diagnostics
PROJECT SUMMARY
U-Sleep is an advanced machine learning model that automates sleep staging—the most time-consuming part of evaluating clinical sleep studies (polysomnography). Developed and validated on data from over 15,000 patients, U-Sleep delivers expert-level accuracy and works across various hardware setups.
CLINICAL IMPACT
U-Sleep automates the time-consuming sleep staging process, which currently takes sleep specialists 2–3 hours per patient. By reducing this burden, U-Sleep enables faster diagnoses, increases patient throughput, and frees clinical resources for treatment and care. This improves access to sleep diagnostics and supports better long-term health outcomes for patients with sleep disorders.