Unobtrusive Long-term Sleep Staging using Photoplethysmography-based Wearables

The diagnosis of sleep disorders is still often based on polysomnography, an in-lab exam allowing experts to perform accurate sleep staging, although this is labor-intensive, expensive, and exposing patients to unusual sleep conditions. Recent work has shown the potential of deep learning models to perform such sleep staging with satisfying accuracy using a simple sensor at the fingertip and the optical technique called photoplethysmography (PPG) commonly used in wearables. In this work, we extended one of these models for the use of PPG data collected from CSEM's wrist-worn wearables in adults and reached encouraging performance (78% accuracy and a Cohen’s kappa of 0.68). Even though the model has shown limitations when applied to patients with cardiac arrythmias (accuracy drop of 10%), it paves the way towards unobtrusive long-term sleep monitoring of patients at home.