A Hierarchical Model Family of Human Gas Exchange for the Use in Medical Decision Support

Abstract

Mechanical ventilation is a life-saving procedure providing the patient with oxygen and allowing sufficient carbon dioxide removal. Although being a routine therapy, finding optimal settings for each patient remains a complex task for the clinician. Mathematical models are a widely accepted tool to simulate physiological processes in the human body, thus being able to predict patient response towards changes in the therapy regime. Such predictions might be exploited to provide medical decision support for the clinician with the goal of finding optimal settings for an individual patient. To allow for suitable reproduction of patient physiology in all clinical situations, rather complex models would be necessary. However, such all-embracing models require an extensive amount of measurements for robust parameter identification without providing further insight into the patient’s disease state. We are therefore proposing to employ multiple model versions each specialized in a different simulation focus and of varying complexity. A medical decision support system might choose from this family of models depending on the patient’s disease state and the physiological parameters that need to be predicted. The proposed model family is hierarchically ordered, i.e. each model is linked to its neighbours. This enables employing model parameters that are identified using simple models to be exploited as initial values for parameter identification in complex models.

Results show, that a simple model of three compartments containing lung and body gas exchange as well as a shunt compartment is suitable for prediction of blood oxygenation for various FiO2 settings. Prediction of carbon dioxide levels at different levels of minute ventilation was achieved with a four compartment model including ventilation/perfusion mismatch that might be extended further by a dead space compartment to simulate tidal breathing. The proposed model family thus demonstrates the practical applicability of hierarchical modelling in clinical practice.