Virtual Intensive Care Unit (ICU): Real-Time Simulation Environment Applying Hybrid Approach Using Dynamic Bayesian Networks and ODEs

Combining deterministic (e.g. differential equations) and probabilistic (Bayesian Networks) approaches to model physiological processes in a real-time software environment leads to a novel model for simulation of human patient physiology especially relevant for intensive care units (ICU). Using dedicated HW/SW interfaces simulated patient signals are measurable with standard monitoring systems. Therefore, this system, based on realistic simulations, is very well suited for teaching and education. Additionally, the environment is usable for inferring patient-specific model structures and parameters. We introduce a hierarchical modeling approach, which allows building complex models based on aggregation of simple sub models. The simulation is controlled to run in real-time with typical sampling times of 1?10 ms (depending on model complexity) on a standard PC (Pentium 2.66 GHz CPU).