Data assimilation in fluid-structure interaction - Application to the cardiovascular system

This joint work with Charles Taylor's Lab at Stanford and the Reo team at Inria Paris-Rocquencourt was initiated with the objective of applying data assimilation strategies to produce patient-specific simulations of a section of the arterial tree. We focused on a healthy patient for which the main difficulty was to introduce and to estimate adequate boundary conditions on the arterial part considered. This work was therefore separated into two contributions. In the first step, we introduced a simple parametrized model of solid boundary conditions representing the external tissue tethering. Indeed, we found that the main limitations of classical fluid-structure interaction simulations for the cardiovascular systems was to neglect the fundamental influence of the surrounding organs around the artery. The model is based on generalized Robin conditions and its parameters have then to be adequately prescribed for a given patient without any standard value in the literature. At this stage, we relied on data assimilation to identify these parameters based on multiple segmentations of a CT sequence registered along a complete cardiac cycle.

This work relies on theoretical studies on observers for fluid-structure interaction systems. The main objective of these studies was to demonstrate the efficiency of the methodology defined in the team for solid mechanics systems to fluid-structure interaction systems. These studies were performed in the context of C. Bertoglio's PhD Thesis from the Reo team in very close collaboration with our team.

Here are the main publications on this subject

And here is an example of the resulting simulation embedded into the CT data