Parameter estimation of transpulmonary mechanics by a nonlinear inertive model.

Transpulmonary mechanics of anesthetized intubated dogs were studied during control breathing and hemorrhage-induced hyperventilation by least-mean-squares parameter estimation using several model versions. The classical elastance-resistance model was modified to include nonlinear elastic and viscous pressure terms with and without a linear inertive pressure component. Inclusion of the nonlinear terms decreased the root-mean-square error of fitting (q) of the classical model on the average to 67% in the control period and to 58% during hyperventilation. An additional decrease due to inertance was 4% (control) and 22% (hyperventilation) and was associated with acceptable estimates of inertance [0.056 +/- 0.02 (SD) and 0.063 +/- 0.008 cmH2O . l-1 . s2, respectively]. When inertance alone was added to the classical model, negligible improvement in q and unrealistic values of inertance were obtained. Conventional measures (Edyn and midvolume resistance) were close to the corresponding least-mean-squares estimates (E and R) of all model versions, except that in hyperventilation neglecting the inertance caused Edyn to markedly overestimate E of nonlinear inertive model.