Coupling Inflammation and Cardiovascular Models for a Sepsis Digital Twin

[zaouache]

Project summary

In this project, I built a simplified digital twin of sepsis by coupling two mathematical models:

• an inflammation model (slow dynamics),

• a cardiovascular model (fast and computationally expensive).

The goal was to link inflammation to cardiovascular response in a simple and efficient way.

What we did

1. Coupling via physiological adapters

The inflammation model does not directly give blood pressure or cardiac output.
We introduced physiological adapters that transform inflammation and tissue damage into hemodynamic factors:

• SVR_scale (systemic vascular resistance),

• contractility_scale (heart contractility).

2. Reduced cardiovascular model (surrogate)

The full cardiovascular model is too expensive to run over long time periods.
We replaced it with a reduced surrogate model that directly predicts:

• MAP from SVR,
• CO from contractility.

3. Fast digital twin simulation

Using this surrogate, we can simulate 200 hours of sepsis evolution without running the full cardiovascular solver.

Key results

The digital twin reproduces physiologically consistent behavior:

• when SVR decreases, MAP decreases (vasoplegia),

• when contractility decreases, cardiac output decreases (myocardial depression).

These results are shown in the figures below.

Conclusion

This work shows that:

• inflammation and cardiovascular models can be efficiently coupled,

• a reduced surrogate can replace an expensive cardiovascular solver,

• long-term sepsis simulations can be performed fast and reliably.