Elements of vision based respiratory monitoring

Traditional respiration sensing in neonatal intensive care often suffers from signal quality issues such as drop-out, presenting difficulties in emerging applications such as automated control of blood oxygen saturation. We are developing a non-contact vision based respiratory monitoring system, suitable for use in a neonatal intensive care unit, to monitor the presence of apnoea, and to provide estimates of respiratory rate and tidal volume as a supplementary system complimenting traditional respiratory monitoring methods. This paper reports on progress relating to data capture and logging and motion detection with this system.

The system comprises a Microsoft Kinect™ webcam and PC running a custom Lab VIEW capture and logging program that takes images of neonates lying in intensive care cribs. Image acquisition methods have been developed allowing capturing of camera images and streaming images to disk at up to 30Hz. Image processing methods haw been developed to automatically identify neonates and to isolate areas of respiratory movement. This basic system will be evaluated in an upcoming pilot study.