Monitoring torso acceleration for estimating the respiratory flow and efforts for sleep apnea detection

Sleep apnea syndrome is a common sleep breathing disorder classified into two major categories: obstructive and central. In this study, we propose a method based on ensemble learning to estimate the respiratory flow, the thoracic respiratory effort and the abdominal respiratory effort from acceleration of suprasternal notch, the thorax and the abdomen respectively. The estimated flow can be used to detect the breathing cessations and the estimated efforts can be used to classify them into obstructive and central apneas. The estimated signals are compared with the signals recorded by welle-stablished measurement methods to show overall mean errors from 11% for the abdomen effort, 17% for the thorax effort and 16% error for the flow estimation. The presented results demonstrate the feasibility of using the torso acceleration as a simple and inexpensive solution for long term measuring and monitoring of respiratory functions for sleep apnea detection.