Assessment of respiratory flow and efforts using upper-body acceleration

In this paper, an innovative method for estimating the respiratory flow and efforts is proposed and evaluated in various postures and flow rates. Three micro electro-mechanical system accelerometers were mounted on the suprasternal notch, thorax and abdomen of subjects in supine, prone and lateral positions to record the upper airway acceleration and the movements of the chest and abdomen wall. The respiratory flow and efforts were estimated from the recorded acceleration signals by applying machine learning methods. To assess the agreement of the estimated signals with the well-established measurement methods, standard error of measurement (SEM) was calculated and $$\rho = 1-{\rm SEM}$$ was estimated for every condition. A significant agreement between the estimated and reference signals was found ( $$\rho = 0.83, 0.82$$ and 0.89 for the estimated flow, thorax and abdomen efforts respectively). Additionally, the agreement of the estimated and reference flows was assessed by calculating the ratio of time at the tidal peak inspiration flow to the inspiration time ( $$t_{\rm PTIF}/t_{\rm I}$$ ) and the ratio of time at the tidal peak expiration flow to the expiration time ( $$t_{\rm PTEF}/t_{\rm E}$$ ). Overall mean and standard deviation of absolute value of differences between $${t}_{{\rm PTIF}}/{t}_{{\rm I}}$$ and $${t}_{{\rm PTEF}}/{t}_{{\rm E}}$$ ratios calculated for every breathing cycle of reference and estimated flow were 0.0035 (0.06) and 0.051 (0.032), respectively. The presented results demonstrate the feasibility of using the upper-body acceleration as a simple solution for long-term monitoring of respiratory features.