BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:Whole-heart electromechanical simulations using latent neural ordi nary differential equations - Matteo Salvador (Stanford University) DTSTART:20240603T103000Z DTEND:20240603T105000Z UID:TALK214480@talks.cam.ac.uk DESCRIPTION:Co-Authors: Marina Strocchi\, Francesco Regazzoni\, Christoph Augustin\, Luca Dede'\, Steven Niederer\, Alfio Quarteroni.\nCardiac digit al twins provide a physics- and physiology-informed framework for predicti ve and personalized medicine. However\, high-fidelity multi-scale and mult i-physics cardiac models remain a barrier to adoption due to their high co mputational cost and the large number of model evaluations required for pa tient-specific personalization. Artificial intelligence-based methods can enable the creation of fast and accurate whole-heart digital twins. We use Latent Neural Ordinary Differential Equations (LNODEs) to learn the tempo ral pressure-volume dynamics of a heart failure patient. Our LNODE-based s urrogate model is trained from 400 3D-0D whole-heart closed-loop electrome chanical simulations\, taking into account 43 model parameters describing cell-to-organ scale cardiac electromechanics and cardiovascular hemodynami cs. The trained system of LNODEs provides a compact and efficient represen tation of the 3D-0D model in a latent space using a feed-forward fully con nected artificial neural network that retains 3 hidden layers with 13 neur ons per layer\, enabling faster than real-time numerical simulations of ca rdiac function on a single processor. This surrogate model is employed to perform global sensitivity analysis and robust parameter estimation with u ncertainty quantification in time frames compatible with clinical practice \, still using a single processor. This framework introduces several compu tational tools for digital twinning in computational cardiology. LOCATION:Seminar Room 1\, Newton Institute END:VEVENT END:VCALENDAR