Kalman Filtering for Posture-Adaptive in-Bed Breathing Rate Monitoring Using Bed-Sheet Pressure Sensors

Breathing rate (BR) is one of the vital signs used in physiological monitoring. Conventional BR monitoring requires attaching wired canula/thermistor on the buco-nasal area to measure air-flow, inducing discomfort to the subject. Abdominal/thoracic belts are also used to detect breathing movements whereas esophageal pressure is the gold standard to measure breathing effort. In this paper, we aim to validate the consistency of using only bed-sheet pressure sensors to monitor the BR in healthy adults. We propose a method and demonstrate that it could be used interchangeably with respiratory belts which were approved for medical use by the American Association of Sleep Medicine (AASM). We build a ten-sinusoidal model-based extended Kalman Filter to adaptively estimate the breathing movements' signal from the body pressure distribution data. The model is posture-specific, I.e., parameters are optimized based on the detected posture. An artificial neural network (ANN) model was used to detect four bed postures to perform the kalman filter parameters' optimization step. The BRs of 12 healthy adults are recorded using the pressure mattress and a reference respiratory belt. To validate the method as a surrogate measure, a Bland-Altman (BA) analysis was performed on both pressure and belt data, and the linear relationship is evaluated using Pearson Correlation Coefficient (PCC). Interestingly, a high inter-rater agreement, an average maximum difference of 1.93 Breaths Per Minute (BrPM), a confidence interval of 95%, along with a strong linear relationship of 95.8% on average between the two methods were interestingly obtained. The presented results show the suitability of the proposed solution in medical applications requiring respiration monitoring.