Efficient computation of interacting model systems

Area of Science:

• Computational physiology and biomedical engineering.
• Development of advanced mathematical models for human physiological systems.

Background:

• Mathematical models can predict human physiological processes for optimizing therapy in medical decision support systems (MDSS).
• Integrating multiple organ system models (respiratory, cardiovascular, gas exchange) for critically ill patients is computationally intensive, limiting MDSS applicability.
• Existing coupled model approaches are too slow for real-time clinical decision support.

Purpose of the Study:

• To develop a computationally efficient method for combining complex physiological submodels.
• To enable the use of integrated physiological models within medical decision support systems for mechanically ventilated patients.
• To reduce the computational cost of complex physiological model simulations.

Main Methods:

• A decoupled computing approach was developed, allowing individual evaluation of respiratory mechanics, cardiovascular dynamics, and gas exchange submodels.
• Interface signals between submodels were estimated and iteratively refined based on model hierarchy.
• Simulation error and time were compared against a traditional coupled computing approach.

Main Results:

• The decoupled approach reduced simulation time by a factor of 34 (one iteration) to 13 (three iterations).
• Maximum simulation error after three iterations was 1.44%, comparable to clinical measurement noise.
• The iterative estimation process converged, minimizing simulation error effectively.

Conclusions:

• The proposed decoupled computing scheme significantly enhances computational efficiency for integrated physiological models.
• This method makes moderately complex physiological model combinations feasible for real-time applications in medical decision support systems.
• The approach provides a viable solution for utilizing complex physiological modeling in critical care settings.