0746 REVIEW OF A MULTISENSOR, LOW COST, AND UNOBTRUSIVE APPROACH TO DETECT MOVEMENTS IN SIT AND SLEEP

Abstract

Introduction: Movement measurements in the Suggested Immobilization Test (SIT) and sleep recordings are typically measured by polysomnography (PSG) with electromyography (EMG). We investigated the viability of an alternate home-based recording system, SleepSmart, which combines sensing technologies integrated in a bedsheet and 3D video to detect movements. Method(s): Pilot study: 19 participants were administered the SIT in the Biomechanics Lab; the subject lay on an angled bed for 30 minutes and slept for up to 90 minutes. We used a combination of the Kinect videography system enabling conventional 2D and novel 3D-technology, a portable EMG device, and a mattress topper sheet fitted with flexible sensors. EMG data was recorded for both tibialis anterior muscles. The goal was to perform pilot testing on the integrated system to fine tune the procedure and equipment. Result(s): Main findings: The 3-D video recordings enabled the study of movement developments, a novel feature not captured by 2-D video- recordings and/or EMG. Pitfalls in the EMG setup, overall protocol design, and data synchronization were encountered. Several requirements were identified to optimize the test-setup: (1) A millisecond-level time stamping system was needed to sync data between multiple modalities; this mechanism will support identification of movement characteristics (development and peak) for Periodic Limb Movements (PLM). (2) Reflective or light-absorbing artifacts should be removed to maintain video data integrity. (3) With the demonstrated effectiveness of the video-data characterization feature, the mattress-sensor framework should implement machine learning algorithms to automatically identify movement events. Conclusion(s): Based on findings, the mattress sensors are being replaced with newer sensors to improve performance. The switch from force-sensing resistors (FSRs) to accelerometers incorporates detection of physiological signals (heartbeat and breathing rate). Identification algorithms will include sleep apnea events. More pilot testing will be conducted to validate changes.