The objective of this thesis is the assessment of a biomechanical heart model using experimental data, and the investigation of clinical applications with patient-specific modeling. At the 1D level we aimed at reproducing physiological experiments with myocardial fiber contraction. For the 3D validation we performed in cooperation with a clinical partner an experiment with animals (pigs) in order to obtain data in the healthy stage and after creating a myocardial infarct. We showed that our model can reproduce the pressures and motion of a healthy heart and that the infarct can be represented by changing only the parameters directly related to the pathology. The objective of the first clinical application was to predict the short-term effect of the Cardiac Resynchronization Therapy (CRT) by means of an increase of 'max LV dp/dt'. The model personalization was performed using patients MRI and pressure data in the baseline condition - prior to CRT. Then we fixed the values of all parameters and applied electrical activation patterns according to the pacing modes considered. We obtained a very good prediction of max LV dp/dt using various pacing patterns in 3 clinical cases. This preliminary clinical validation shows that the modeling of CRT is a very promising approach as an assistance to therapy planning. The second application is based on the adaptation of data assimilation methods developed in the MACS team at INRIA. We performed joint state-parameter estimation with real image data. We showed the effectiveness of these algorithms in automatic model personalization and that the estimated contractility values can serve as an indicator of the local heart function.
