BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:HIGH-SPEED HIGH-FIDELITY CARDIAC SIMULATIONS USING A NEURAL NETWOR
 K FINITE ELEMENT APPROACH - Michael Sacks (University of Texas at Austin)
DTSTART:20230804T080000Z
DTEND:20230804T090000Z
UID:TALK202465@talks.cam.ac.uk
DESCRIPTION:A comprehensive image-based computational modelling pipeline i
 s required for high-fidelity patient-specific cardiac simulations. However
 \, traditional simulation methods are a limitation in these approaches due
  to their prohibitively slow speeds. We developed a physics-based training
  scheme using differentiable finite elements to compute the residual force
  vector of the governing PDE\, which is then minimized to find the optimal
  network parameters. We used neural networks for their representation powe
 r\, and finite elements for defining the problem domain\, specifying the b
 oundary conditions\, and performing numerical integrations. We incorporate
 d spatially varying fiber structures into a prolate spheroidal model of th
 e left ventricle. A Fung-type material model including active contraction 
 was used. We developed two versions of our model\, one was trained on a re
 duced basis of the solution space\, and one was trained on the full soluti
 on space. The models were trained against two pressure-volume loops and va
 lidated on a third loop (Fig. 1). We validated our implementation against 
 conventional FEM simulation using FEniCS. While the reduced order model wa
 s trained faster than the full-order model\, we achieved mean and standard
  deviation of the nodal error between the NNFE solution and the FE solutio
 n with 10 -3 cm\, with both models\, where the characteristic length was 1
  cm (Fig. 2a). The NNFE model predicted each solution within 0.6 ms wherea
 s the FE models took up to 500 ms for each state. The NNFE method can be s
 imultaneously trained over the entire range of physiological boundary cond
 itions. The trained NNFE can predict stress&ndash\;strain responses for an
 y physiological boundary condition without retraining.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR