08:30
MS35: Large-scale Biomechanics: Models, Solvers and HPC (Part 1)
Chair: Lorenzo Zanon
08:30
25 mins
|
Towards a Digital Human: An Overview
Lorenzo Zanon, Nehzat Emamy, Andreas Hessenthaler, Thomas Klotz, Aaron Krämer, Benjamin Maier, Tobias Rau, Thomas Ertl, Dominik Göddeke, Michael Krone, Miriam Mehl, Oliver Röhrle
Abstract: Biophysically detailed multi-scale models, such as the chemo-electro-mechanical skeletal muscle model proposed in [1], enable realistic simulations of biological systems. These can be often accomplished only by means of large-scale compute facilities. In [2], the simulation of the contraction of skeletal muscle tissue has been upgraded from a moderately parallel version towards an efficient, massively parallel one, yet not sufficiently detailed for an in-depth understanding of the neuro-muscular system. In this talk, we present further advances in this endeavour, as an outcome of the project “DiHu” (funded by the “Baden-Württemberg-Stiftung”).
In [3], our simulated model for the action potential propagation in a biceps brachii has been extended to a realistic number of fibers of 270.000. An appropriate domain decomposition allows us to run simulations on 27.000 cores of the supercomputer HazelHen at the HLRS in Stuttgart, Germany. Configurability, efficient data structures and modular software architecture of the utilised open-source softwares target usability, performance and extensibility for future models. We investigate choices of linear solvers and preconditioners including a multigrid method, and also illustrate the efficient use of MPI in order to reach good parallel scaling.
In conjunction with parallel scalability, another key target is the efficient use of modern hardware resources. We design robust and flexible code and code-generators that efficiently exploit the available hardware in a problem-independent way. Our main focus lies in an optimized kernel distribution among CPUs and – when appropriate – GPUs, governed by directive-based programming models – such as OpenMP 4.5.
In the visualization framework MegaMol, we also exploit CPUs and GPUs for efficient rendering of the large-scale-simulated muscle fibers. Tools such as the I/O framework ADIOS2 apply a minimal overhead to the intra-node communication using shared memory. In particular, data can be processed in-situ right during the simulation with no need of storing large amounts of simulation results.
REFERENCES
[1] T. Heidlauf and O. Röhrle, “A multiscale chemo-electro-mechanical skeletal muscle model to analyze muscle contraction and force generation for different muscle fiber arrangements”, Frontiers in Physiology, 5, 498 (2014).
[2] C. P. Bradley, N. Emamy, T. Ertl, D. Göddeke, A. Hessenthaler, T. Klotz, A. Krämer, M. Krone, B. Maier, M. Mehl, T. Rau, and O. Röhrle, “Enabling Detailed, Biophysics-based Skeletal Muscle Models on HPC Systems”, Frontiers in Physiology, 9, 816 (2018).
[3] B. Maier , N. Emamy, A. Krämer, M. Mehl, “Highly Parallel Multi-Physics Simulation of Muscular Activation and EMG”, Proceedings in Coupled Problems (2019).
|
08:55
25 mins
|
An efficient time discretisation of the incompressible elastodynamics equation in living tissues
Federica Caforio, Sébastien Imperiale
Abstract: The principal aim of this work is to provide an adapted numerical scheme for the approximation of elastic wave propagation in living tissues.
We rely on high-order conforming finite element with mass lumping for space discretisation and implicit/explicit, second-order, energy-preserving time discretisation.
The time step restriction only depends on the shear wave velocity and at each time step a Poisson problem must be solved to account for the incompressibility constraint that is imposed by penalisation techniques. To do so, we use a fast solver adapted to high order finite elements.
|
09:20
25 mins
|
Reduced Basis Methods for Real-Time Thermal Ablation Cancer Treatment Planning
Zoi Tokoutsi, Martin Grepl, Karen Veroy, Marco Baragona, Ralph Maessen
Abstract: Patient specific treatment planning determines the placement of multiple ablation probes and the power control of the ablation device based on a desired tissue temperature increase. We consider two strategies to solve this problem.
The first one is a multi-step planning algorithm which involves which involves the solution of a partial differential equation constrained optimal control problem and a subsequent greedy placement of the ablation devices. In the second approach we first parametrize the problem with respect to uncertain parameters of interest, e.g. tissue properties, geometric parameters and then employ the reduced basis method to construct a real-time efficient, reliable surrogate model. We use this model to efficiently solve a bilevel optimal control problem for the optimal placement parameters. We present numerical results for both strategies and compare the results.
|
09:45
25 mins
|
Cardiac Fluid-Structure Interaction
Boyce Griffith, Marshall Davey, Robert Hunt, Simone Rossi, Margaret Anne Smith, David Wells, Charles Puelz
Abstract: Fluid-structure interaction (FSI) is a fundamental aspect of cardiac mechanics, particularly in the function of the four cardiac valves. The immersed boundary (IB) method [1] is an approach to fluid-structure interaction that was introduced by Peskin to model the fluid dynamics of heart valves. This talk will present work that aims to develop a medical imaging-derived IB model of the heart, its valves, and the nearby great vessels that includes biomechanical models of the major cardiac structures that, wherever possible, uses experimental data from human tissue specimens. The talk will describe IB methods that are amenable to the use of material descriptions that use the framework of large deformation nonlinear mechanics [2], verification studies that demonstrate that these methods can yield accuracy that is comparable to specialized methods for incompressible nonlinear elasticity [3], and validation studies that focus on comparisons to experimental data obtained in an in vitro experimental model of the heart [4]. The talk also will describe work to accelerate these computations through load balancing strategies that are tailored to IB methods. The talk also will present recent large-scale simulations of the human heart that leverage these methods.
References:
1. C.S. Peskin. The immersed boundary method. Acta Numerica, 11:479–517, 2002.
2. B.E. Griffith and X.Y. Luo. Hybrid finite difference/finite element version of the immersed boundary method. International Journal for Numerical Methods in Biomedical Engineering, 33(11):e2888, 2017.
3. B. Vadala-Roth, S. Rossi, and B.E. Griffith. Stabilization approaches for the hyperelastic immersed boundary method for problems of large-deformation incompressible elasticity. arXiv preprint arXiv:1811.06620.
4. J.H. Lee, A.D. Rygg, E.M. Kolahdouz, S. Rossi, S.M. Retta, N. Duraiswamy, L.N. Scotten, B.A. Craven, and B.E. Griffith. Fluid-structure interaction models of an experimental pulse duplicator for simulating bioprosthetic heart valve dynamics. engrXiv preprint 8ys2k.
|
|
