Parameter Estimation & Uncertainty Quantification

Team: Mitchel Colebank, Amanda Colunga, & Mette Olufsen
Collaborators: Dirk Husmeier & Mihaela Paun

Overview: Computer models of physiological systems are becoming increasingly important in understanding complex biological phenomenon. In particular, model parameters within these models can serve as a surrogate for true biomarkers, informing experimental design and exposing underlying physiological mechanisms. These models are usually nonlinear with numerous parameters, hence identification of parameters and their corresponding uncertainty is nontrivial. Using a combination of sensitivity analysis and subset selection, an influential, identifiable subset of parameters can be determined, reducing model complexity. In addition, it is crucial to quantify the uncertainty in the output of a model given the variabilities in the inputs and parameters. Both frequentist and Bayesian methods can be used to quantify confidence and predictions intervals for a given quantity of interest.

To this extent, we seek to understand: (i) efficient methods for sensitivity analysis and subset selection, (ii) how to infer model parameters from limited, noisy physiological data, and (iii) uncertainty propagation for cardiovascular models.


Click here to return to our general research page!