Optimized estimation of respiratory impedance by signal averaging in the time domain

The spontaneous breathing of a subject during measurements of respiratory impedance (Zrs) by the forced oscillation technique (FOT) induces errors that result in biased impedance estimates, especially at low frequencies. Although in standard measurements this bias may be avoided by using special impedance estimators, there are two applications of FOT for which such estimators are not useful: when a head generator is used and when measurements are made during intubation. In this paper we describe a data-processing procedure for unbiased impedance estimation for all FOT setups. The proposed estimator (Z) was devised for pseudorandom excitation and is based on time-domain signal averaging before frequency analysis. The performance of estimator Z was first analyzed by computer simulation of a head generator setup and a setup including an endotracheal tube to measure (2-32 Hz) a resistance-inertance-elastance model mimicking Zrs of a healthy subject. Second, Z was assessed during real measurements in 16 healthy subjects. The results obtained in the simulation (e.g., error in elastance was reduced from 15.6% with most conventional estimators to 3.3% with Z in simulation of head generator setup) and in the measurements in subjects (differences of less than 1.6% between Z and a reference) confirmed the theoretical lack of bias of Z and its practical suitability for the different FOT setups. In addition to its applicability in the situations in which no other unbiased estimators are available, estimator Z is also advantageous in most conventional applications of FOT, since it requires much less computing time and thus allows on-line Zrs measurements.