Recursive identification of lung parameters

Abstract

Determination of lung capacity (FRC) using insoluble gas washout or equilibrium methods is a common procedure in respiratory tests. The lung model can be extended to include multiple compartments with differing volumes and ventilation fractions. A discrete-time mathematical model of a multi-compartment lung was developed based on mass conservation laws of the gas exchange. To estimate the unknown parameters in the model from experimental data, a recursive prediction error algorithm was implemented. The design of the algorithm provides the capability to track time-varying parameters, such that the system can be monitored on-line. Determination of the model order may be as important as estimation of the parameters, especially in biological systems where little may be known about the structure. Two new criteria to distinguish between models of varying complexity, the ‘minimum description length principle’ and the ‘accumulated prediction error’, are illustrated. Theoretical studies show that these criteria yield consistent order estimates of the process. Simulations and experimental data confirm the feasibility of this approach for the estimation of lung parameters.