Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Explore how physics-informed neural networks can create patient-specific digital twins for cardiac hemodynamics in this 47-minute conference talk from Stanford University's MedAI Group Exchange Sessions. Learn about a novel approach that treats digital twin development as a composite inverse problem, drawing parallels to self-supervised learning's pretraining and fine-tuning methodology. Discover how researchers pretrain neural networks to learn differentiable simulators of cardiac physiological processes, then fine-tune these models to reconstruct physiological measurements from echocardiogram videos while maintaining physical equation constraints. Understand the potential for eliminating invasive procedures by using non-invasive patient data to create personalized medical simulations that can predict treatment outcomes and guide clinical decisions. Examine the intersection of machine learning and domain physics knowledge in building interpretable models for cardiovascular health, particularly focusing on applications in cancer survivorship and long-term cardiovascular outcomes. Gain insights into computational precision health methodologies that address challenges with unobserved or unobservable medical data through physics-informed approaches.
