Efficient sensitivity analysis of models with many model parameters to guide model personalization

Sensitivity indices obtained from a global variance-based sensitivity analysis can help by identifying which parameters need to be measured accurately and which parameters can be fixed for patientspecific applications. However, the computational cost for obtaining the sensitivity indices becomes prohibitively large with the current gold standard approach when the number of model parameters becomes too high (> 25). We propose an efficient two-step approach for model personalization. Using this approach personalization of an existing model could be performed using O(103) model runs, compared to O(105) runs had the current gold standard approach been used