Fully-Automatic Camera-Based Pulse-Oximetry During Sleep

Tom Vogels, Mark van Gastel, Wenjin Wang, Gerard de Haan; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 1349-1357

Current routines for the monitoring of sleep require many sensors attached to the patient during a nocturnal observational study, limiting mobility and causing stress and discomfort. Cameras have shown promise in the remote monitoring of pulse rate, respiration and oxygen saturation, which potentially allows a reduction in the number of sensors. Applying these techniques in a sleep setting is challenging, as it is unknown upfront which portion of the skin will be visible, there is no unique skin-color outside the visible range, and the pulsatility is low in infrared. We present a fully-automatic living tissue detection method to enable continuous monitoring of pulse rate and oxygen saturation during sleep. The system is validated on a dataset where various typical sleep scenarios have been simulated. Results show the proposed method to outperform the current state-of-the-art, especially for the estimation of oxygen saturation.