[home] [Personal Program] [Help]

---

Session: Poster session II
Session starts: Thursday 24 January, 16:00

---

Axel Riedlinger ()
Jörn Kretschmer ()
Knut Möller ()

Abstract:
Patient-specific mathematical models effectively predict physiological processes. They might be exploited to support therapeutic decision making directly at the bedside. In mechanically ventilated patients, the oxygenation status of the patient should at all times be sufficient. Gas exchange models are able to describe concentrations and distribution of both oxygen (O2) and carbon dioxide (CO2) inside the lung and the blood circuit. To represent a specific patient, model parameters need to be adapted to his individual behavior via measurements from the patient monitoring. Models of gas exchange with different complexity and therefore with different accuracy are described in the literature [1, 2]. The more complex the model, the higher the number of model parameters that have to be adapted to mimic a specific patient. Selecting suboptimal initial values often leads to model parameters deviating far from their correct values. In models with numerous parameters, fitting results may strongly depend on the chosen initial values because of the high number of possible parameter combinations. Thus, parameter identification of complex models with numerous parameters is potentially not robust. Erroneous model parameters would lead to a model that does not represent the patient status adequately leading to an unfavorable therapy recommendation. A hierarchical model structure may support robust identification of model parameters and allow flexible simulations of patient’s reactions to therapeutic decisions [3]. We have built a hierarchically structured family of gas exchange models, where the simplest gas exchange model acts on the assumption of a single lung compartment plus a shunt describing an area of the lung that is perfused but not ventilated. This model can be exploited to estimate the shunt parameter using a minimal set of information. This allows the reduction of the amount of parameters to be identified in the more complex models where different ventilation and perfusion rates (V/Q mismatch) are implemented [4]. Applying such a hierarchical parameter identification approach therefore simplifies model adaptation, making it more robust. First results of parameter identification of complex gas exchange models are promising. Further verification by means of patient data is necessary to prove this finding.