Parameter Identification in Models of Respiratory Mechanics – A Hierarchical Approach

Abstract

Models of respiratory mechanics predict certain effects of ventilator settings on the patient’s lung in order to reduce potential harmful effects of the ventilation therapy. As the prediction quality improves with the degree of individualization, parameter identification is used to determine the parameter constellation that represents the lung properties best. The complexity of parameter identification increases with the complexity of the employed model i.e. the number of tunable parameters. To keep the identification process feasible, a hierarchical approach was developed.

This approach is exemplarily illustrated by identifying the viscoelastic model (VEM) of respiratory mechanics. The results of the identification of simpler models are incorporated in order to support the subsequent parameter identification of the following more complex model.

Advantages of the hierarchical approach over the common direct approach are illustrated regarding the decoupled influence of the initial estimates and the reduced computing time using simulated and clinical data.

The underlying investigations proved that a hierarchical approach is beneficial with respect to robustness and efficiency.