Oleg Portniaguine and Dmitriy Pavlov

1
Oleg Portniaguine and Dmitriy Pavlov
Scientific Computing and Imaging
Institute University of Utah
Convenient Modeling of Bioelectric Fields
1
SCI
2
Outline
Introduction Computational Scheme Numerical
Results Conclusion
2
SCI
3
Introduction
EEG Modeling Boundary Elements Finite
Elements Voxel grid based on MRI (Bonovas et al,
2001 Schinpf et al, 1998) - supercomputer needed
3
SCI
4
Introduction ...
We propose a computational scheme based on MRI
voxel size cubic grid with ILU preconditioner to
solve EEG problem on PC in a matter of
seconds The modeling domain consists of regular
cubic elements voxels with the same size as
underlying MRI or CT 3-D image of the object
4
SCI
5
Computational Scheme
Scalar electric field potential within the voxel
satisfies the Laplas-type equation ?????
0 and can be approximated by a trilinear
function ????a1xyza2xya3xza4yza5xa6ya7za8
Assuming ? const within the voxel a1 0
5
SCI
6
Element Template
?2
?02
?03?
?04?
?0
?3
?4
?01?
?1
6
SCI
7
Scheme Coefficients
Current continuity contitions at the voxel edges
sn(?n - ?0n) s0(?0n - ?00)
Coefficient equation
7
SCI
8
ILU preconditioned BiCJ relaxation
A?? b A DUL-U, D diag(A) L(DUL-U)U-1U?
???Lb L(DLU-1-I)m c The Bi-conjugate gradient
algoritm is used to find m
8
SCI
9
Conductive Sphere Validation
One dipole Grid size 128x128x128 (2 mm MRI)
9
SCI
10
Realistic Head Model
One dipole Grid size 112x128x128 (2 mm MRI)
10
SCI
11
Conclusion
The modeling code shows its reliability and
fastness, while more validation is needed The
code can be easily modified for anisotropic
conductivity case The code will be modified to
simulate magnetic field as well for the MEG
modeling The ultimate goal is to use this
computational scheme as a part of fast and
accurate inversion code for EEG/MEG and other
source localization problems
11
SCI
12
Acknowledgments
This project is supported by NIH/NCRR
grant 1-P41-RR12553-01A1
12
SCI