This work addresses major challenges of heart model personalization. Analysis techniques for clinical intracardiac electrograms determine wave direction and conduction velocity from single beats. Electrophysiological measurements are simulated to validate the models. Uncertainties in tissue conductivities impact on simulated ECGs. A minimal model of cardiac myocytes is adapted to the atria. This makes personalized cardiac models a promising technique to improve treatment of atrial arrhythmias.

This cumulative dissertation emphasizes the advantages of regression models (in general) as well as of Bayesian statistics (compared to frequentist statistics), introduces a newly developed graphical user interface (GUI) for fitting Bayesian regression models in a convenient way, and demonstrates how predictions from a submodel resulting from the (Bayesian) projection-predictive feature selection (PPFS) can be performed easily in clinical practice (despite the comparatively complex computations behind these predictions).eng