Global sensitivity analysis for assessing the parameters importance and setting a stopping criterion in a biomedical inverse problem

This paper shows how to obtain in addition the standard deviations available after a data assimilation procedure based on ensemble Kalman filter, an apportioning of total uncertainty outputs patient-specific blood flow model into small portions due input parameters. Statistical indicators generally used for identifying importance numerical parameters, namely Sobol' first order and indices, are introduced discussed. These allow identification rank different parameters model, as well influence interactions between these output variance. The results show that knowing during is useful avoid unnecessary over-solving find suitable stopping criterion clinical situations where faster diagnosis always requested. Indeed, work permits reduce typically by factor six time solution most importantly with very limited extra calculation using already information.