Proceedings of the 2005 IEEE

Engineering in Medicine and Biology 27th Annual Conference

Shanghai, China, September 1-4, 2005

A Fundamental Study on Parameter Estimation

of Layerd Local Tissue Impedance for EIT
Emiko Yasuno, Hiromi Kato, Yohsuke Kinouchi and Tadaoki Morimoto

Abstract— Electrical Impedance Tomography (hereinafter re- subcutaneous structure [17], [18]. They focused on the local
ferred to as EIT) is 2-D or 3-D image of electrical impedance tissue, and measurement electrode and bio-impedance space
distribution in a living tissue. Unlike usual imaging methods, distribution estimation has been studied.
i.e., Xray-CT, MRI and US Imagings, EIT is used for imaging
the information of tissue structure and functions. This paper In our recent studies, a new configuration of the electrode,
provides a new estimation method as a fundamental study to called ”divided electrode”, was proposed for a short time
realize a EIT for local biological tissue. Up until now we has measurement of bio-impedance in a cross section of a local
proposed a new configuration of the electrodes, called divided tissue. The cross section of the tissue was represented by
electrode, for a high-speed measurement of bio-impedance in space distributed equivalent circuits, and their parameters
a cross section of a local tissue. The cross section of the tissue
was represented by space distributed equivalent circuits of were estimated by inverse algorithm. However, we needed
tissue structure known, and their parameters were estimated the shape of the tissue to grasp with US imaging, etc.
by inverse algorithm. In this paper, we try to estimate the In this paper, it was confirmed to be able to estimate the
parameter value of a layered structural model, the thickness of parameter value of a layered structural model, the thickness
the layer, and the boundary without using US-imaging by using of the layer, and the boundary without using US-imaging by
the divided electrode. Its capability is examined by computer
simulations, where a distributed equivalent circuit is used as using the proposed divided electrode.
a model of the tissue. Estimation of impedance parameter is
carried out by use of the Gauss-Newton method. Usefulness
II. METHOD
of the proposed method is confirmed by computer simulations A. Divided Electrode
using a typical layered tissue model.
Figure 1 shows the principle for noninvasive measurement
I. INTRODUCTION of local tissue impedance in a cross section using divided
At present, Xray-CT, MRI and US-imagings, which show electrode. Theoretically, the divided electrode is the same as
the density distribution and density boundary information, the four electrodes method. The divided electrode has the
are widely used for medical tissue diagnosis. On the other shape of plate divided by slits and consists of many voltage
hand, EIT has focused as a new method to obtain a tissue electrodes, current electrodes, and guard electrodes. Figure
structures, and physiological functions and states of the tissue 1(a) shows top view of the divided electrode. As show in
[1], [2]. It has obtained current and voltage measurements Fig.1(b), the voltage electrodes arranged at the center are
around the surface of a body as a result of applied a divided into B parts (v1 , v2 , · · · , vB ). The current electrodes
weak alternating current. The measurements are analyzed by arranged on both sides of the voltage electrodes are also
various data restructuring algorithms, and they express it as divided into A parts (i1 , i2 , · · · , iA ). The guard electrodes are
2-D or 3-D images of the electric impedance distribution in arranged at the top and bottom with both sides (iA+1 , iA+2 )
a living tissue. So far bio-impedance information has been of the current electrodes. Currents flow from all the currents
applied to macro measurement of lung functions[3], blood and the guard electrodes at the same time. At this time,
circulation functions[4], the amount of body composition[5], the currents flow into the section S of the center without
the amount of body fat, etc. Though many reserches have spreading as shown in Fig.1(c), because guard currents hold
been reported[6], [7], [8], [9], [10], [11], [12] to realize the down the measurement current from the both sides. For
EIT for practical use, there remain several problems to be example, denoting the current into j-th current electrode by
solved. The problems are inverse ploblem algorithm for esti- ij , the potential of k-th voltage electrode by vk and Fourier
mating parameters, and electrode structure and arrangement transform by Ij and Vk , measured impedance Zj,k is defined
for practical use. by:
We examined usefulness of total bio-impedance for diag- Vk
Zj,k = (j = 1, · · · , A, k = 1, · · · , B) (1)
nosis of tumor [13], [14], [15], [16] and for estimation of Ij
E. Yasuno is with the Department of Systems and Control Engineering, That is, measured impedance Zj,k represents around inter-
Anan National College of Technology, 265 Aoki Minobayashi, Anan, section of current line of flow ij and isoelectric line of
Tokushima 774-0017, Japan [email protected]
H. Kato and Y. Kinouchi are with the Department of Electrical and Elec- the circuit model vk . By use of the measurement data of
tronic Engineering, Faculty of Engineering, The University of Tokushima, currents (i1 , · · · , iA ) and voltages (v1 , · · · , vB ), C(= A ×
2-1 Minamijosanjima-cho, Tokushima 770-8506, Japan B) impedance data are obtained at once. Hence, the 2-
T. Morimoto is with the School of Health Sciences, Faculty of Medicine,
The University of Tokushima, 3-18-5 Kuramoto-cho, Tokushima 770-8503, D impedance distribution can be obtained[12]. Moreover,
Japan configuration of current electrodes controls the measurement

0-7803-8740-6/05/$20.00 ©2005 IEEE. 6650

area. The voltage electrodes control measurement area in the On the other hand, lumped circuit is used as another
electrodes-axis direction. Therefore, unlike previous multi method. It originates in intracellular resistance, extracellular
electrodes method, divided electrode is realized high-speed resistance, and cell membrane capacitance. The lumped
and high-resolution measurement. circuit can simply represent biological tissue structure. How-
ever, in the case of three parameters equivalent circuit, it has
only one relaxation time and it expresses a biological tissue
Voltage source va in an equivalent circuit on the tissue level. But lumped circuit
model doesn’t include non-uniformity of the tissue. Then
v1 vk vB
iA i1 i1 iA spatial distributed equivalent circuit is used as the model
2 2
considered on the cell level. This is considered that each cell
is expressed in an equivalent circuit, much they gather, and
the biological tissue is constituted. In the case of a uniform
living tissue, an equivalent circuit on tissue level is express
Voltage electrode Current electrode Guard electrode by intracellular and extracellular resistances Ri , Re and cell
membrane capacitance Cm in β dispersion.
(a) Top view Moreover, the cross section of layered tissue as shown
in Fig.3 can be expressed in the 2-D distributed equivalent
circuit. In this model, three parameters circuits are connected
Slit
in the shape of a lattice. The electrodes are point electrodes.

Voltage electrode Current electrode Guard electrode

(b) Bottom view

Voltage source va
Current
Voltage electrode electrode
v1 vk vB Guard
electrode
iA 2 i j i1 i1 ij iA 2

Fig. 2. Two dimensional distributed equivalent circuit model.

A cross section Z jk, C. Estimation method

of the tissue S
As mentioned above, impedance data ZD of C are mea-
sured by the divided electrode consists of current electrodes
(c) Side view of A part, voltage electrodes of B part. Moreover, if measure-
ment frequency are set with ωj (j = 1 ∼ U ), it describe the
Fig. 1. Divided electrode. following an estimate method of each parameter. Measured
impedance ZD can be expressed as following:
⎡ ⎤
ZD (1) (ω1 )
B. Distributed equivalent circuit model ⎢ .. ⎥
⎢ ⎥
⎢ . ⎥
Although functions of a living tissues are the same for ⎢ (1) ⎥
⎢ ZD (ωU ) ⎥
each tissue, non-uniformity of a tissue exists as to com- ZD = ⎢ (2) ⎥ (2)
⎢ ZD (ω1 ) ⎥
ponents and dimensions. Usually, this non-uniformity is ⎢ ⎥
⎢ .. ⎥
expressed by two methods. One of them is assuming that ⎣ . ⎦
the distribution of electric relaxation time(time constant of ZD (C) (ωU )
circuit model) is Cole-Cole distribution[19] or Davidson- On the other hand, the model connected to N equivalent
Cole distribution[20]. Cole-Cole circle rule using the Cole- circuits of three parameters in Fig.3 is considered. Parameter
Cole distribution is suitable for expressing impedance of vector p is expressed as:
the large area of beta distribution region. However, it is
difficult to express the average characteristic of the living p = [Re (1), Ri (1), Cm (1), · · ·
tissue separately. · · · , Re (N ), Ri (N ), Cm (N )]T (3)
6651
where T is transposed matrix. And circuit model impedance
ZM is expressed as:
⎡ (1) ⎤
ZM (p, ω1 )
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ (1) ⎥
⎢ Z (p, ωU ) ⎥
ZM (p) = ⎢ M (2) ⎥ (4)
⎢ ZM (p, ω1 ) ⎥
⎢ ⎥
⎢ .. ⎥
⎣ . ⎦
(C)
ZM (p, ωU )
The error ε(p) of circuit model impedance ZM and
impedance data ZD is defined by:
ε (p) = ZM (p) − ZD (5)
To get the solution ε(p) which equals 0, the Gauss-Newton
method is used, i.e., δp is the change in parameter is given Fig. 3. Layerd circuit model.
by:
 

∂ε p(t) TABLE I
− δp(t) = ε p(t) (6)
∂p PARAMETER VALUE OF THE MODEL AND INITIAL VALUES .

then,
Model parameters p Initial parameters p0
Re Ri Cm Re Ri Cm
Xδp = Y (7) (Ω) (Ω) (nF) (Ω) (Ω) (nF)
 T
∂ε(p(t) , ω1 ) ∂ε(p(t) , ωU ) Layer1 110 110 9 70 70 15
X = ,···, (8)
∂p ∂p Layer2 110 110 9 70 70 15
 T Layer3 50 50 34 70 70 15
T T Layer4 50 50 34 70 70 15
(t) (t)
Y = ε(p , ω1 ) , · · · , ε(p , ωU ) (9) Layer5 50 50 34 70 70 15
Layer6 83 83 16 70 70 15
Layer7 83 83 16 70 70 15
X is U · C × 3N matrix and Y is U -dimensional vector Layer8 83 83 16 70 70 15
here. X can be obtained from the numerical analysis of the Layer9 83 83 16 70 70 15
circuit in Fig.3. Therefore, calculation of the least squares Layer10 83 83 16 70 70 15
method of Eq.(8) expressed with Eq.(10) obtained δp, which
is parameter change from the initial model corresponding to
measurement impedance. figure is not considered. Each parameter are estimated from
−1  100 data using initial values as shown in TableI.
T T T T
δp = X X + X X X Y+X Y (10) The parameter values are estimated correctly. The relative
error of estimated parameters is shown in TableII, and the
III. RESULTS error curve is shown in Fig.4.
The usefulness of the proposed method has been verified
by computer simulations using layered tissue model. Because
measurement area of this electrode is local tissue, breast
is modeled here. Actual tissue is in the order of the skin,
fat, the mammary gland, and muscle. Layer structure model
composed of fat, the mammary gland, and the muscle as
shown in Fig.3 is used because of the skin is thin when it
sees relatively.
An unknown parameter is 30 because it assumes that each
parameter value is also the same in the same layer though Fig. 4. Estimated result (Error curve).
the number of circuit elements is 409. Five current electrodes
(from i1 to i5 ) and one voltage electrode are arranged at
equal intervals, respectively. Measurement frequency points IV. CONCLUSIONS AND FUTURE WORKS
are 20 in range of 0 to 150kHz. At this time, the number of In this paper the bio-impedance in consideration of un-
measurement data is 5 × 1 × 20=100. Here, because tissue uniformity of tissue is investigated for the purpose of realiz-
in the direction of the electrode axis is uniform in layer ing EIT. In addition, 2-D distributed equivalent circuits com-
structure model, the guard electrode located in the center part posed of intracellular and extracellular resistances Ri , Re
of right and left both edges of measured electrode shown in and cell membrance capacitance Cm was considered. As the
6652
 TABLE II
[8] J. J. Ackmann and M. A. Seitz,”Methods of complex impedance
R ELATIVE ERROR OF ESTIMATED PARAMETERS . measurements in biologic tissue”, Crit. Rev. Biomed. Eng., 11, pp.
281-311, 1984.
Relative error(%) [9] N. Chauveau, B. Ayeva, B. Rigaud and J. P. Morucci, ”A multifre-
Re (Ω) Ri (Ω) Cm (nF) quency serial EIT system”, Physiol. Meas, 17, pp. A7-A13, 1996.
[10] Y. Yamamoto and T. Yamamoto,”Measurement of electrical
Layer1 0 0.0091 0 bioimpedance and its applications”, Medical Progress through
Layer2 0.0091 0.0091 0.0011 Technology, 12, pp. 171-183, 1987.
Layer3 0 0.28 0 [11] H. Kato, E. Yasuno, Y. Kinouchi, and T. Morimoto, ”Electrocal
Layer4 0.02 0.08 0 Impedance Tomography for Local Biological Tissue”, Proc. of the
Layer5 0 0 0 8th International Conference on Control, Automation, Robotics and
Layer6 0.012 0.012 0.063 Vision, pp. 942-946, 2004.
Layer7 0.012 0.024 0 [12] X. Zhao, Y. Kinouchi, T. Iritani, T. Morimoto and M.
Layer8 0.012 0.048 0 Takeuchi,”Estimation of Multi-Layer Tissue Conductivities from
Layer9 0 0 0.13 Non-invasively Measured Bioresistances Using Divided Electrodes”,,
Layer10 0 0.048 0 IEICE TRANS. INF. & SYST., Vol.E85-D, No.6, pp. 1031-1038,
2002.
[13] T. Morimoto, Y. Kinouchi, T. Iritani, et al.，”Measurement of the Elec-
trical Bio-impedance of Breast Tumors”, European Surgical Research,
measuring method suitable for practical use, using divided Vol. 22, pp. 86-92, 1990.
[14] T. Morimoto, S. Kimura, Y. Konishi, et al., ”A study of the electrical
electrodes has been proposed. In order to estimate the bio- bioimpedance of tumors”, J. Invest. Surg., 6, pp. 25-32, 1993.
impedance spatial distribution, parameters of the equivalent [15] S. Kimura, T. Morimoto, T. Uyama, et al., ”Application of electrical
circuits have been estimated by use of the Gauss-Newton impedance analysis for diagnosis of a pulmonary mass”, Chest, 105,
pp. 1679-1680, 1994.
method. As a consequence, it is found that high precision [16] Y. Konishi, T.morimoto. Y. Kinouchi, T. Iritani and Y. moden,
parameter estimation was possible for the complicated model ”Electrical Properties of Extracted Rat Liver Tissue”, Research in
by use of a divided electrode. Moreover, the method has been Experimental Medicine, Vol. 195, pp. 183-192, 1995.
[17] N. Momose, Y. Kinouchi, T. Iritani, et al., ”Three dimensional analysis
applicable also to the model of thickness and the boundary of local tissue impedance by a finite element method”, World Congress
in the tissue unknown. on Medical Physics and Biomedical Engineering, pp. 347, 1997.
According to the results of the evaluation computer sim- [18] A. Ihara, T. Morimoto, X. Zhao, Y. Kinouchi, T. Iritani and Y.
Oomine，”Finite Element Analysis of Breast Bioimpedance”, Proceed-
ulations, it is found from that the proposed method is very ings of The 4th Asia-Pacific Conference on Medical and Biological
useful and practical as a new EIT technology. Engineering, pp. 452, 1999.
Future works will improve the accuracy of the model’s [19] K. R. Foster and H. P. Schwan, ”Dielectrical properties of tissues and
biological materials : a critical review”, Critical Reviews in Biomedical
spatial resolution, and develop the measurement electrode Engineering, 17, pp.25-104, 1989.
using divided electrode. [20] D. W. Davidson and R. H. Cole,”Dielectric relaxation in glycerol,
propylene glycol, and n-propanol”, J. Chem. Phys., 19, pp.1484-1490,
1951.
V. ACKNOWLEDGMENTS
This research was supported in party by the Grant-in-Aid
of Encouragement of Young Scientists for JSPS, Japan.

R EFERENCES
[1] L. A. Geddes and H. E. Hoff, ”The measurement of physiologic events
by electrical impedance,” a review, Am. J. Med. Electr., Jan. -March,
pp. 16-27, 1964.
[2] Y. Kinouchi, T. Iritani, T. Morimoto and S.Ohyama, ”Fast in vivo
measurements of local tissue impedances using needle electrodes”,
Medical & Biological Engineering & Computing, pp.486-492, 1997.
[3] L. E. Baker, ”Applications of the impedance technique to the respira-
tory system”, IEEE Eng. Med. & Biol. Mag., 8, pp. 50-52, 1989.
[4] R. P. Patterson，”Fundamentals of impedance cardiography”, IEEE
Eng. Med. & Biol. Mag., 8, pp. 35-38, 1989.
[5] K. Sakamoto, K. Kaneko, M. Ezaki, et al., ”Estimation of human
wholes body fat volume by electrical impedance method”, Jpn. J. Med.
Electro. Biol. Eng., 33, pp. 2-9, 1995.
[6] Y.Yamashita et al.,”Method and feasibility of estimating impedance
distribution in the human torso”, Proc. 5th ICEBI, pp. 87, 1981.
[7] K.Nakayama et al., ”Fundamental study on electrical CT algorithm
utilizing sensitivity theorem on impedance plethysmography”, Proc.
5th ICEBI, pp. 99, 1981.

6653