Effect of generator nonlinearities on the accuracy of respiratory impedance measurements by forced oscillation.

Measurements of respiratory impedance by means of the forced oscillation technique (FOT) are usually made using a loudspeaker as the excitation device. Its nonlinear nature can introduce artifacts that coincide with the frequencies applied to excite the respiratory system, limiting the accuracy of the impedance estimation. In this paper, this hypothesis is evaluated in the case of both a traditional estimator and the unbiased estimator proposed by Daróczy and Hantos (1982). A simulated study under apnoea conditions in the pressure range 0.5-3.0 cmH2O peak-to-peak reveals that loudspeaker nonlinearities introduce a characteristic pattern of dispersion in both the resistance and reactance curves that can be significantly decreased (p approximately equal to 0.03, signtest) by reducing the nonlinearities. A simulation of spontaneous breathing shows the same pattern, and is observed in the case of traditional as well as unbiased estimators. The dispersion is quantified by the mean absolute distance between the theoretical and simulated data and decreases with the reduction of nonlinearities when impedance is estimated with a traditional estimator (from 6.63 to 4.72% in real estimates and from 6.78 to 3.47% in imaginary estimates) as well as with an unbiased estimator (real estimates from 4.84 to 1.57% and 5.61 to 2.06% in imaginary estimates). Studies with normal subjects show the same dispersion pattern, which decreases if the generator nonlinearities are reduced. These results supply substantial evidence that reducing generator nonlinearities can contribute to the production of more reliable mechanical impedance FOT measurements.