Personalizing computational models to construct medical digital twins.

Area of Science:

• Biomedical engineering
• Computational biology
• Digital health

Background:

• Digital twin technology, initially for engineering, is now applied to biomedicine.
• Personalizing computational models with patient data is crucial for predictive healthcare.
• Current methods struggle to bridge macroscale clinical data with microscale model requirements.

Purpose of the Study:

• To develop a novel algorithm for dynamically calibrating patient-specific computational models.
• To address the challenge of linking macrostate (clinical data) and microstate (model parameters) in personalized medicine.
• To improve the accuracy of model-based forecasts in healthcare applications.

Main Methods:

• Developed an algorithm applying the ensemble Kalman filter at the macrostate level.
• Linked macrostate updates to microstate updates for agent-based models.
• Ensured microstates are compatible with macrostates and likely according to model dynamics.

Main Results:

• The algorithm successfully bridges the gap between clinical measurements and fine-grained model data.
• Generated microstates compatible with desired macrostates and model dynamics.
• Provides a novel approach for personalizing complex biomedical models.

Conclusions:

• The developed algorithm offers a robust method for personalizing agent-based models in healthcare.
• This approach facilitates more accurate predictions for personalized medicine.
• Enables better utilization of patient data for dynamic model calibration.