The purpose of our work is to personnalize an atrial model of the propagation of the action potential, based on electrical catheter data. We use sequential data assimilation with a state observer model and a Kalmann filter. The state observer is a Luenberger observer of the atrial model. It can pursue the actual activation front reconstructed from cathether data. The Kalman filter is designed to estimate the electrical conductivity coefficients. To this aim, our cardiac solver CEPS has been coupled to the data assimilation library Verdandi, developped by the Inria team M3disim. The method has been evaluated against analytical data computed from a direct monodomain solver.

Afterwards, we have been confronted firstly to CARTO data from a patient from Bordeaux's hospital. In this case, the signals need a lot of preprocessing to be usable, and there are missing information on a large part of the atria. The model need tuning to be used in this context, but allows to recover the missing information, and complete the activation front on the whole atria. Anyway, the dataset is insufficiant to evaluate the methodology. Secondly, we have been recently working on richer Rythmia data from two patients. The dataset consists in a segmented atrial volume and electroanatomical maps in sinus rhythm, for a single pacing site, and an atrial arrhythmia. For this dataset, we registered the Rythmia geometry to the segmentation, we prepared the computational mesh, defined some relevant anatomical regions, fiber directions, and projected the Rythmia data on the segmented model. We are currently running our first data assimilation computations.

This is a joint work with Annabelle Collin (MONC Research Team, INRIA), Jason Bayer (IHU Liryc), Antonio Frontera (IHU Liryc), Gautier Bureau (M3DISIM Research Team, Inria), Philippe Moireau (M3DISIM Research Team, INRIA) and Yves Coudière (CARMEN Research Team, INRIA).