Data-driven computational models of heart anatomy, mechanics and hemodynamics: An integrated framework

Cardiac therapies aim to correct pathological blood flow. Patient-specific therapy planning is challenging due to the large variability in disease cause, location and severity. A predictive framework is therefore needed to assess the optimal treatment for a patient in terms of maximizing effectiveness (blood flow velocity, vorticity, cardiac output, etc.) and minimizing the risk of complications. Multi-physics fluid structure interaction models have been proposed to investigate the function of the bi-ventricular myocardium. Yet, no or only partial patient-specific data have been used due to the lack of efficient methods for accurate modeling of patient's cardiac anatomy from images. Current imaging technologies enabling assessment of cardiac anatomy, dynamics and tissue structure, we propose an integrated framework for multi-physics heart modeling based on imaging data. Our approach relies on efficient machine learning methods to estimate an accurate and comprehensive model of patient's anatomy from MRI. A computational model of cardiac electrophysiology and biomechanics is then solved on patient's anatomy.