Collaborative Research: SCH: Machine-Learning-Enhanced Computational Models of Cardiac Pathophysiology

Computational models enhanced with machine learning have the potential to significantly improve our ability to simulate biological processes at multiple length scales. This includes various disease conditions that affect organ structure and function. One such disorder is pathological fibrosis, which is the excess accumulation of extracellular matrix (stiff fibers) within the organ tissue. While this process is compensatory initially, prolonged fibrotic activity can cause increased stiffness leading to dysfunction. This project combines techniques and insights from several disciplines, including engineering, computer science, applied mathematics, and physiology, to develop advanced computational models of the heart. The work aims to build fundamental understanding of heart-disease progression and could aid in the evaluation of potential treatment strategies. Since cardiovascular disease is the leading cause of death in the United States, discoveries from this research could have a significant impact on society. Additional broader-impact aspects of the work include incorporation of technical examples of cardiac engineering into computational-mechanics courses along with the development of open-source software tools and databases. The goal of this research is to develop a computationally-efficient multiscale modeling framework that integrates machine learning and artificial intelligence to predict the structural and functional changes that occur in the presence of heart disease. Specifically, it will create a multiscale finite element framework that uses innovative computational techniques to couple (1) network models (myocardial perfusion), (2) agent-based models (myocardial fibrosis), and (3) timescale separation schemes (myofiber growth). All of these developments integrate concepts from mathematics and engineering to define two-way interactions between system-level biology and molecular mechanisms. To enhance this computational framework even further, a physics informed neural networks will be developed to provide instantaneous calculations of these complex interactions, via efficient replication of the multiscale framework output. This approach will leverage emerging ideas from computer science and applied mathematics. Finally, the computational tool kit will enable a deeper understanding of how microstructural changes in tissue composition affect whole heart function, as well as the effects of potential therapies. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.