Abstract
This thesis follows the development of a clinically useful, meshless algorithm for
simulation of soft tissue. The algorithm is designed to operate in irregular 3D
geometry and allow for automatic discretising. It employs fully nonlinear geometric
and material formulations and handles almost incompressible media. It includes
multiple parts and contacts while being fast and sufficiently accurate on consumer
hardware.
Before developing the algorithm, existing commercial software (LS-DYNA) is
used to perform some simulations of craniotomy-induced brain shift. The results
suggest that the Element Free Galerkin method will be a useful foundation for
future simulations.
New software is written to implement a non-conforming background grid instead
of a structured element mesh for quadrature. Errors introduced by the use of a
non-conforming background grid, are an order of magnitude smaller than normal
surgical precision. Simulations are performed with arbitrarily placed nodes to allow
for irregular geometries, but this often introduces errors and instability. So efficient
relationships are developed between node density, shape function support size,
explicit time step size, and the non-conforming quadrature grid.
Simple geometric problem domains are used to verify the method against an
established Finite Element solver. Errors are deemed negligible in the context of
clinical applications. The increased flexibility in discretisation and the capacity to
handle extreme deformations, are considered more valuable than the small loss of
accuracy.
Indentation of porcine brains is simulated and compared to experimental data.
The results are accurate enough to pursue the method in a more complete surgical
simulation. So a complete simulation of craniotomy-induced brain shift involving
irregular geometry, nonlinear formulations, arbitrary discretisation, multiple parts,
multiple materials, and a contact algorithm is performed. The results demonstrate
that realtime simulations are possible on existing consumer hardware.
In order to improve efficiency, we consider combining the meshless method with
a suitable Finite Element method. Preliminary tests are performed on cylinders
and ellipsoids to confirm that the combination functions correctly. In a simulation
of tumour growth the combined method is found to be more efficient than a pure
meshless method and more stable than a pure Finite Element method. The union
of these two methods yields a fast, stable, and efficient simulation.
It is concluded that the meshless method has potential for clinical use in surgical
simulation.
simulation of soft tissue. The algorithm is designed to operate in irregular 3D
geometry and allow for automatic discretising. It employs fully nonlinear geometric
and material formulations and handles almost incompressible media. It includes
multiple parts and contacts while being fast and sufficiently accurate on consumer
hardware.
Before developing the algorithm, existing commercial software (LS-DYNA) is
used to perform some simulations of craniotomy-induced brain shift. The results
suggest that the Element Free Galerkin method will be a useful foundation for
future simulations.
New software is written to implement a non-conforming background grid instead
of a structured element mesh for quadrature. Errors introduced by the use of a
non-conforming background grid, are an order of magnitude smaller than normal
surgical precision. Simulations are performed with arbitrarily placed nodes to allow
for irregular geometries, but this often introduces errors and instability. So efficient
relationships are developed between node density, shape function support size,
explicit time step size, and the non-conforming quadrature grid.
Simple geometric problem domains are used to verify the method against an
established Finite Element solver. Errors are deemed negligible in the context of
clinical applications. The increased flexibility in discretisation and the capacity to
handle extreme deformations, are considered more valuable than the small loss of
accuracy.
Indentation of porcine brains is simulated and compared to experimental data.
The results are accurate enough to pursue the method in a more complete surgical
simulation. So a complete simulation of craniotomy-induced brain shift involving
irregular geometry, nonlinear formulations, arbitrary discretisation, multiple parts,
multiple materials, and a contact algorithm is performed. The results demonstrate
that realtime simulations are possible on existing consumer hardware.
In order to improve efficiency, we consider combining the meshless method with
a suitable Finite Element method. Preliminary tests are performed on cylinders
and ellipsoids to confirm that the combination functions correctly. In a simulation
of tumour growth the combined method is found to be more efficient than a pure
meshless method and more stable than a pure Finite Element method. The union
of these two methods yields a fast, stable, and efficient simulation.
It is concluded that the meshless method has potential for clinical use in surgical
simulation.
| Original language | English |
|---|---|
| Qualification | Doctor of Philosophy |
| Supervisors/Advisors |
|
| Publication status | Unpublished - 2015 |