P004 A novel method for nightly body position tracking during sleep

Abstract Introduction Body position likely contributes to substantial night-to-night variability in obstructive sleep apnoea (OSA) and snoring. However, non-contact methods for multi-night body position assessment are lacking. This study is investigating the feasibility and performance of a novel non-contact mattress-sensor for continuous body position monitoring during sleep. Methods This ongoing in-laboratory study involves healthy adult volunteers getting in- and out-of-bed and performing replicate body position changes (supine, left, right, and prone) whilst laying on a custom-built mattress-sensor strip containing a piezoelectric and three pressure sensors. Participants also wear a tri-axial accelerometer on their chest for body angle measures. In- and out-of-bed events are determined from the presence versus absence of (1) body-weight via the summed pressure sensor signals, and (2) cardiorespiratory motion via the standard deviation of the piezoelectric signal. A Generalized Additive Model (GAM) is used to fit pressure sensor data to accelerometer roll angles and classify supine versus non-supine positions. Results Data from the first three participants comprised 15 in- and out-of-bed events and 115 body position changes. All in and out-of-bed events were correctly detected with a mean (SD) timing difference of 2.1 (±1.1) seconds. The GAM model predicted continuous roll angles with an adjusted R2 of 0.73 (77.7% deviance explained) and area under the ROC curve for supine versus non-supine position classification of 0.92 (95% CI: 0.86–0.96). Conclusions These preliminary data support the feasibility and potential utility of a non-contact mattress sensor system for nightly monitoring of body position effects on OSA and snoring.