With the advances of modern medicine decision making for physicians is a complex task involving the predicion of nonlinear time variant processes. Due to the probabilistic nature of medical reasoning characterized by incomplete and noisy data most decision support systems in medicine are either rule based systems or probabilistic reasoning approaches e.g. Bayesian networks. Besides those a different path is taken by a number of researchers who incorporate physiological models into decision making. Major problems that have to be dealt with: 1. creating adequate models during the reasoning process, 2. identification or individualization of models at the bedside, 3. active generation of data informative enough for parameter estimation.
Hierarchical models to support intensivists in ventilation support will be discussed.

Knut Möller studied Computer Science with application to medicine at Bonn University, Germany and Rutgers UniversityNJ. He was a visiting scholar at the Instituto Leonardo Fibonacci Trento, Italy, before he received a Ph.D. degree in Computer Science and later an M.D. in Medicine at Bonn University, Germany. Currently, he is the Dean of the Biomedical Engineering Division at Furtwangen University, Germany, associated professor at the Clinics for Anaesthesiology and Intensive Care Medicine, University Hospital of Freiburg, Germany and Co-Director of the Institute for Technical Medicine (ITeM) Freiburg, Germany.

His research interests include physiological modelling, intelligent signal analysis, decision support systems and automation with special focus on intensive care medicine and rehabilitation engineering.