[home] [Personal Program] [Help]

---

Session: Medical Instruments - Surgery II
Session starts: Friday 25 January, 13:00
Presentation starts: 13:15
Room: Lecture room 557

---

Christoph Schranz (Furtwangen University)
Knut Möller (Furtwangen University)

Abstract:
The application of respiratory mechanics models combined with standardized ventilation manoeuvres enable investigations of patients’ lung mechanics directly at the bedside to optimize ventilation therapy [1]. Therefore, underlying effects of respiratory mechanics such as viscoelasticity, inhomogeneity and recruitment are uncovered by applying certain flow profiles. Subsequently, these effects can be captured by the corresponding model via parameter identification methods. Data sets of 13 patients with Acute Respiratory Distress Syndrome (ARDS) undergoing various ventilation maneuvers [2] are available along with a hierarchical model structure for simulation and parameter identification. The hierarchy is structured according to the model complexity. The first hierarchical layer contains the basic linear 1st order model (FOM) consisting of a serial arrangement of airway resistance and a compliant lung compartment. The second layer contains extensions of the FOM, linear 2nd order models (SOM) assuming viscoelastic tissue contributions (VEM – Viscoelastic Model) or ventilation inhomogeneities (IHM – Inhomogeneity Model). Additionally, the 2nd layer also includes model extensions of the FOM whith either a nonlineare resistance or compliant element. In the given hierarchy, a pressure dependent compliance model (PRM) is implemented [3]. The PRM was developed to simulate the sudden opening of alveolar regions (recruitment) by exceeding a certain opening pressure as observed in ARDS patients [4]. The third hierarchical layer contains a combination of the VEM and PRM, where viscoelasticity is assigned to the recruitable alveolar regions. All models in a particular layer are extensions of the simpler models in the layer above. Thus, the hierarchical structure can support gradient-based parameter identification processes by offering convenient initial values via a prior identification of simpler models in the hierarchy [5]. The applicability of a patient-specific FOM in various ventilation maneuvers proved to be critical since distinct time-depending effects are not considered. In contrast, a individualized SOM captured the observed dynamics and provided accurate simulations under static and dynamic conditions. Pressure dependent effects were successfully captured by the PRM. However, due to the lack of model-dynamics the patient-specific PRM was only applicable in the corresponding maneuver. Finally, only individualized PRVEMs could capture pressure depending recruitment effects together with the observed dynamics to provide accurate simulations in various scenarios. The proposed model hierarchy illustrates the applicability of different models under various circumstances and enables robust parameter identification of more complex models. Due to the combination of the models in the 2nd layer a first suggestion of a respiratory mechanics model could be proposed that is able to provide accurate simulations of a patient in various conditions.