μ-Finite Element Analysis of Human Bone Structures

Prof. Dr. P. Arbenz, ETH Zurich, Institute of Computational Science
Prof. Dr. G. H. van Lenthe, KU Leuven, Department of Mechanical Engineering
Prof. Dr. R. Müller, ETH Zurich, Institute of Biomechanics
Dr. M. Sala, ETH Zurich, Institute of Computational Science (until Nov 30, 2006)

Keywords: high performance computing, bone structure analysis, osteoporosis

Funding: This project is supported by grant #200021-113950 of the Swiss National Science Foundation.

Abstract

Osteoporosis is a disease characterized by low bone mass and deterioration of bone microarchitecture. It leads to increased bone fragility and risk of fracture, particularly of the hip, spine and wrist. According to the WHO, lifetime risk for osteoporotic fractures in women is close to 40%, in men about 13%. Osteoporosis is second only to cardiovascular disease as a leading health care problem. Osteoporotic fractures are a major cause of severe long-term pain and physical disability, and have an enormous impact on individual, society and health care systems.

With the advent of fast and powerful computers, simulation techniques are becoming popular for investigating the mechanical properties of bone. Using microstructural finite element (μFE) models generated directly from computer reconstructions of trabecular bone it is now possible to perform a 'virtual experiment', i.e. to simulate a mechanical test in great detail and with high precision.

Ideally, the development of a system with microstructural resolution below 50 micron would allow in-vivo measurement of patients at different instances in time and at different anatomical sites. Unfortunately, such systems are not yet available, but the resolution at peripheral sites in now below 100 micron, a level that allows elucidation of individual microstructural bone elements.

The resulting μFE models are computationally demanding and require special solution schemes. The preconditioning conjugate gradient method is the obvious solution method. However, currently used preconditioners are in general not effective, and their quality deteriorates as the problem sizes become large.

The goal of the proposed research is two-fold.

  1. We will rewrite the code that is presently used. The new code shall be fully parallel, employing an advanced algebraic multigrid preconditioner and a parallel file system. With the advent of sufficiently large computers (e.g. at the Swiss National Supercomputing Center (CSCS) in Manno) it will become possible to solve systems of equations of up to one billion degrees of freedom. This will enable us to calculate physiological tissue loading in whole bones, which is of great importance in the assessment of the mechanical competence and fracture risk of these bones.
  2. A second memory efficient code shall be developed that will allow solving systems of equations of the size that are today solved routinely on supercomputers on multiprocessors with 16 to 32 cores. It is assumed that such machines will be the most widely used parallel machines in five years time and available at a cost that a research group can afford.

Media Appearance

Publications

Software

Contact

Prof. Dr. Peter Arbenz
Institute of Computational Science
Universitätsstrasse 6, CAB G69.3
CH-8092 Zürich
Tel.: +41 44 632 74 32
Email: arbenz@inf.ethz.ch

Last update: 10 November 2008