Documente Academic
Documente Profesional
Documente Cultură
0.5
0.5
1.5
1.5 1 0.5 0 0.5 1 1.5
Figure 1: A typical sensor geometry indicating radial fins, screen and vessel wall
4.1 Introduction
Electrical Capacitance Tomography (ECT) attempts to image to permittivity in the
interior of an object from external capacitance measurement. Mathematically it is
the inverse conductivity (EIT or ERT) problem. The electric potential satisfies
0
The discrete version of the Dirichlet to Neumann mapping is the trans-capacitance
on each electrode and Q is the vector Qk
matrix C, which satisfies Q CV , where V is a vector of the potentials specified
Ek n dS (where Ek is the kth
1
1.5
0.5
0.5
1.5
1.5 1 0.5 0 0.5 1 1.5
Figure 2: A typical finite element mesh for the ECT sensor generated using the
EIDORS ECT package and QMG
electrode). The trans-capacitance C Ci Ce is the sum of the interior and any
wall of the pipe Ci , and known and fixed exterior capacitance Ce .
Permittivities of commonly encountered materials vary by a factor of about 3,
and in many practical situations there are known upper and lower bounds. This
means that the well known logarithmic stability estimates apply.
2
4.3 Regularized solutions
We repressent the permittivity as a piecewise constand function on a square or
triangular mesh. The Jacobian J dC d is readilly calculated
ij
Jk
Ci j k i j dA
Pk
where Pk is the kth mesh element (pixel),k the permittivity of that pixel and i is
the potential when electrode i is driven.
For a single linear step, or a difference image between two measurements of
capacitance, we need a regularized solution to
J C
The traditional approach to this taken in ECT is to use Landwebers iteration as
a regularized inversion procedure [3]. It is interesting that the first iteration of
Landweber, which has rather confusingly been called linear back projection in
the ECT literature, can sometimes produce useful images.
Landwebers algorithm for any number of iterations, like Tikhonov or any
filter of the singular value decomposition (see Bertero and Boccacci [7]) can be
implemented to run almost as fast as one iteration of Landweber, once the SVD
of J has been calculated. However the projected Landweber method [7] incorpo-
rating prior upper and lower bounds on has been found to work well in ECT,
and has no such acceleration. Fast imaging is often desirable in ECT as the data
collection rate can be for example 50 frames per second for a 12 electrode system.
A reconstruction of an object covering six neighbouring triangles, and the
graphical user interface for the EIDORS ECT system is shown in Figure 5.3.
When there are larger permittivity contrasts, an example being mixtures of
oil, air and water where the ratio may be as high as 30, nonlinear reconstruction
methods must be employed. The approach we have taken is to minimize
C C 2
2 L 2
JiT Ji 2 LT L
i JiT C C 2 LT Li
3
1
1.5
0.95
0.9
0.5
0.85
0
0.8
0.5
0.75
1
0.7
1.5
1.5 1 0.5 0 0.5 1 1.5
Figure 3: The graphical user interface for EIDORS ECT, showing a reconstruction
from simulated data
References
[1] M. Vauhkonen, W.R.B. Lionheart, L.M. Heikkinen, P.J. Vauhkonen and J.P.
Kaipio, A Matlab Toolbox for the EIDORS project to reconstruct two- and
4
three-dimensional EIT images, Proceedings of Conference on Biomedical
Applications of Electrical Impedance Tomography, University College Lon-
don, April 5 - 7, 2000. See also
!#" .
[5] G Alessandrini, E Rosset, JK Seo, Optimal size estimates for the inverse
conductivity problem with one measurement, Proc. Amer Math. Soc 2000,
Vol.128, pp.53-64
[6] W.Q. Yang and M. Byars, An Improved Normalisation Approach for Electri-
cal Capacitance Tomography, Proc 1st World Congress on Industrial Process
Tomography, Buxton, Derbyshire, 1999, pp215-218