Synopsis

Cardiac Magnetic Resonance offers a multitude of contrast mechanisms across spatiotemporal scales to further improve diagnosis, stratification and interventions in heart failure patients. In this presentation the focus is put on hyperpolarized metabolic and diffusion tensor imaging to unravel the interplay of energy supply and myofiber contractile reserve in conjunction with biophysical modeling of left-ventricular function and dysfunction. It is shown that statistical learning and, in particular, physics-informed neural networks offer approaches for rapid personalization, detection and prediction of failing hearts.