Efficient probabilistic model personalization integrating uncertainty on data and parameters: Application to...

Area of Science:

• Computational physiology
• Biophysical modeling
• Medical applications

Background:

• Biophysical models are vital for organ-scale medical applications but lack confidence measures due to data sparsity, noise, and soft tissue complexity.
• Uncertainty quantification is critical for the clinical utility of personalized computational physiology models.
• Estimating patient-specific parameters for stochastic models remains a challenge.

Purpose of the Study:

• To develop an efficient Bayesian inference method for personalizing biophysical models.
• To enable uncertainty quantification in patient-specific computational physiology.
• To facilitate the clinical application of personalized models by providing confidence measures.

Main Methods:

• Utilized polynomial chaos and compressed sensing for efficient Bayesian inference.
• Applied the method to cardiac electrophysiology models.
• Validated on synthetic data and applied to clinical data for uncertainty assessment.

Main Results:

• Demonstrated the feasibility of Bayesian inference for 3D modeling problems.
• Quantified the impact of data characteristics on model personalization and prediction accuracy.
• Showcased the method's ability to handle uncertainties in data and model parameters.

Conclusions:

• The proposed method makes Bayesian inference computationally tractable for complex 3D models.
• This approach provides essential uncertainty quantification for personalized biophysical models in clinical settings.
• The method supports the optimization of treatment by providing confidence in personalized model predictions.