Accuracy in the noninvasive estimation of pulmonary blood flow by identification of the CO2 uptake system

Estimation of pulmonary blood flow, controlled by a roller pump, was carried out in dogs using a system identification technique. The system parameters in a linearized CO2 uptake model were estimated by the maximum-likelihood method from the input consisting of alveolar ventilation and inspired pCO2 and the output of end-tidal pCO2. The auto- and cross-correlations of the residuals showed that the model used for estimation was appropriate. The pulmonary blood flow calculated from the system parameters contained a degree of error which showed an average coefficient of variation of 23.8%. Error analysis revealed that the main sources of error lay in the estimates and in the error-magnifying effect arising in the calculation of the flow from these estimates. Reduction of the error in the estimates requires a long set of data for estimation; on the other hand, the error-magnifying effect is determined by physiological constants, so that it cannot be small even in clinical application.