Escolar Documentos
Profissional Documentos
Cultura Documentos
Thesis
Extended abstract
1. Introduction
From the cardiologist point of view the classical 12-lead ECG is a simple, low cost
diagnostic tool. However the limited numbers of sampling points (9 active electrodes
spread over the entire body) and the way to display this information (time varying
signals) make the analysis process extreme complicated. The information content is low
comparing to the complexity of the 3-dimensional heart electrical phenomena. The
transfer from the spatial, time-varying heart potential field into a limited set of 12 signals
hides a lot of useful information. Long time training of the cardiologist for easy
recognition of typical waveform distortions, huge data bases for various heart diseases,
and sophisticated automatic analysis and diagnostic algorithms, make this tool
valuable. A further step forward, in order to acquire more primary information, is the
Body Surface Potential Mapping - BSPM technique. These systems use 32-90
electrodes to sample the thorax surface and build potential maps at relevant moments
within the cardiac cycle. The clinical use of these maps depends on the physician ability
to interpret them. Although a greater information content compared to classical ECG,
the BSP maps are yet hard to understand. The propagation process from the heart
surface to the thorax surface, makes the two potential fields total different, and the lack
of clinical expertise needed to interpret the BSPM leads to limited extend of its clinical
use. Other BSPM displaying methods as QRS area maps, ST segment maps, activation
time maps have limited application in diagnostic of the myocardial infraction, ectopic
beats, WPW syndrome or intra-vetricular blocks. A difficult point with both classical ECG
and BSPM is the mental effort needed to associate the heart electrical events with the
recorded patterns on the thorax. Under these circumstances a simpler way to see
directly whats happen on the heart surface is a more attractive concept.
The activation front concept
One of the most popular model for the electrical heart activity is the activation front [3].
The myocard cells are activated one by other or by the specialised nervous tissues of
the heart, change the electrical state and became negative (depolarisation process).
That part of the heart muscle that is already activated is negative related to the not yet
activated part. A potential step of 40 mV appears over this 0.5 mm wide separation
zone. This potential step travels across the myocard as more of the heart became
activated. The myocard depolarisation process begins at the upper side of the right
atrium in the vicinity of the sino-atrial node (SNA) and travels down and left thought the
right and left atrium. After a short delay, the activation front breaks out again near the
apex and then travels upwards toward the hearts base (see Fig.1). Maps with the time
evolution of this activation front where invasively recorded using epicardial electrodes
into opened thorax by Scher (1956), Durrer (1970) and Spach (1975). The ability to
1
Ph.D. Thesis
Extended abstract
SAN
PR
AVN
HISS
Ph.D. Thesis
Extended abstract
Still an electrical model for the medium between the heart and body surface is needed.
We used a simplified thorax model in order to keep the calculation's complexity low. The
internal structure between thorax surface and
heart surface is built from a finite number of
entities, bounded by closed surface and having
an electrical conductivity approximately
constant in their volume. With a certain degree
of accuracy, these entities can be associated
with various anatomic parts of the thorax
(bones, lungs, liver, spinal cord, kidney...).
Useful prior information about the geometry
(electrode position, thorax shape and organ
size and position) could be measured using an
imaging method like CT, MRI or ultrasound
tomography. A better division of the thorax and
Figure 2 Electrical model of the a greater number of homogeneous volumes,
thorax. The actual electrodes give a better precision for the inverse solution.
position depends upon the patient The glue medium that fills the space between
the homogeneous volumes is the thorax volume
thorax topology.
VT with the mean conductivity T and bounded
outside the thorax by the surface ST and inside by the heart surface SH and the surface
of the homogeneous entities Si (i=1..M). Therefore VT does not includes the heart
volume VH neither the volumes Vi. The 2nd Green equation applied to this volume
gives:
VT ( )dV
S T ( n T n T )dST S H ( n T n T )dS H
M
i 1
(1)
S i ( n T n T )dSi
S i ( n i n i )dSi 0
(2)
Replacing (2) in (1) and considering the continuity of the potential and its normal
derivate at entity surface, results the final potential equation:
Ph.D. Thesis
Extended abstract
ST n T dS T SH n T n T dS H
M
1
dS i 0
T Si n T
i 1
(3)
Equation (3) can be transformed from continue to discrete by digitising the surfaces. We
choose a number of N discrete nodes for every surface of interest (thorax, heart and
entities. By using a numeric integration method, the integrals are now finite sums over N
points of the numerical values in nodes, and (3) is now:
N
( rTj )
j1
n T
Tj
N
( r Hj )
Hj
Hj
( r Hj )
n
n
T
T
j1
N
( rij )
0
k i (Sij)
T
i 1
j1
M
(4)
with ki = 1-i/T the conductivity step at the Vi interfaces. We can use this equation in
order to find the unknown electric potential on the heart surface. Additionally we have
as supplementary unknowns the potential on the entity's surface and its normal
derivative on the heart surface up to a total number of N(M+2) unknowns. In order to
solve for these unknowns, we have to write N(M+2) equations for N(M+2) different ro
positions. A simple approach is to let ro to travel through the N(M+2) digitising nodes.
However these nodes are on the boundary, not outside VT. The Boundary Elements
Method chose the position of ro inside the VT volume in points very close to the nodes.
However we found the mathematical and numerical problems associated with the huge
Green function values in points ro near the integral surface difficult to handle. Therefore
we chose these positions outside VT uniformly spread over a cylinder, 1m height, 1m
diameter, centred around the thorax in order to avoid the singularity of the Green
function and its derivative in the nodes. Equation (4) builds a linear system of N(M+2)
algebraic equations with N(M+2) unknowns, which allow us to solve for the potential and
normal current on the heart surface.
Ph.D. Thesis
Extended abstract
or muscle somestesic stimulation. For the reconstruction process most of the authors
use the Finite Element Method [5,6] in order to avoid the fact that the internal geometry
of the thorax is initially unknown. Therefore the number of initial unknowns is bigger as
the number of finite elements used to describe the thorax is greater. The reconstruction
process is very difficult especially for the 3-dimensional case [5,6]. To simplify the
complexity of the calculus we used the Boundary Elements Methods and an identical
thorax model with the one used for the inverse problem in cardiology. In this case we
have to find only the conductivity of the homogeneous entities that build-up the thorax.
In order to apply the reconstruction algorithm one needs to know the transfer
relationship which gives the resulted thorax voltages produced by the injected currents.
In the frequency range used for exploration, tissues are mainly resistive and the
electromagnetic field produced inside the thorax is stationary harmonic. We wrote the
Green equation using effective values for currents and voltages. Similar with (4) the
following conditions holds true: inside the thorax there are no sources for the voltage
(=0) and =0 because ro VT and the current normal to thorax surface is null
everywhere except electrode's surface. We agree also that the electrodes are small
compared to ro and hence is constant over one electrodes surface. Equation (3) will
have the new form:
N
S T n T dST I j ( rej )
j 1
M 1
1 i
dS i 0
T S i n T
i 1
(5)
where Ij is the current, rej is the position vector of the jth electrode and SM+1 SH.
The discrete version when considering a number of N nodes for each surface will be:
N
( rTj )
j 1
n T
Tj
M 1
I j ( rej )
j 1
( rij )
0
k i Sij
T
i 1 j 1
N
(6)
Equation (6) allows us to build a consistent system of equations for the total number of
U = N(N-1)/2 independent current patterns and resulted voltages. Hence U is the
maximum number of unknown conductivity values possible to find when using N
exploring electrodes. Similar to the IPC we have to deal with N(M+1) additionally
unknowns for each test pattern, associated with the resulted potentials in the
homogeneous entities' nodes. We handle this problem by sweeping ro position over the
above mentioned cylinder.
Ph.D. Thesis
Extended abstract
Multiplexed
isolation amplifier
Multiplexed
isolation amplifier
Multiplexed
isolation amplifier
Multiplexed
isolation amplifier
R1
C1
C2
A1
A2
R2
R4
-0,6V
D1
D2
R3
R5
2x
DZ
C3
Ph.D. Thesis
Extended abstract
R1
A1
R1
A2
C1
R
C2
Vref
A2
R2
R4
R1
A3
D1
D2
R3
R5
2x
DZ
C3
Cc
A4
MUX
8
canale
ADG
508
Rg
Vref
UWilson x 100
C2
Demod
factor
de umplere
Ph.D. Thesis
Extended abstract
wandering synchronous to the respiration movements that are due to a poor electrode
contact. The constant base line shifting is eliminated based on a statistical procedure.
The histogram of the signal discrete amplitudes is used to find the deviation from null of
the base line. A linear interpolation procedure eliminates the time aperture error
introduced by the delays between the sampling moments at each electrode appearing
due the use of a single AD converter.
For both IPC and EIT problems we used a thorax model with 5 homogeneous entities
(lungs, liver, stomach and spinal cord) each digitised with 32 nodes.
In the first attempt we used typical values for
the electrical parameters measured in vitro"
by Kim [2]:
Tissue type
Thorax - mean
Heart muscle
Lungs
Spinal cord
Liver
Stomach
[.cm]
300
450
1200
2000
625
800
T T ( r0 ) H H ( r0 ) H H ( r0 ) I I ( r0 ) 0
(7)
I 1
We used this equation as descried above to build a linear system of equation that we
solved to find the unknowns associated with the potential on the heart surface. Because
of the ill-posed nature of the problem we had to use a Tikhonov first order regularisation
procedure to keep the inverse solution stable.
Figures 5 and 6 show the evolution of the signals measured on the thorax of a healthy
29-year old male subject and the reconstructed signals on the heart surface.
Additionally the analysis software has the possibility to display an animated image that
reproduces at reduced speed the evolution of potential on heart surface by coding the
potential values as different grey levels or different colours.
Ph.D. Thesis
Extended abstract
Figure 8 The ECG signals recorded on the torso surface (Wilson Central
Reference) for a young healthy patient
Ph.D. Thesis
Extended abstract
4
0.189
0.811
0.433
1.056
1.678
1.678
0.433
0.433
1.056
1.433
2.056
0.811
2.678
0.189
2.056
1.433
1
proof of the Helmholtz theorem for equivalent sources in 2 cases: unipolar current
layer and bipolar current layer
proof of the reciprocal relationship between thorax surface potential distribution and
heart surface potential distribution in the case of a non-homogenous thorax volume
situated into a bounded region
the original method of injected currents used to proof the reciprocal relationship
between thorax potential and heart surface potential
deducing of the discrete transfer matrix between the potential measured on the
thorax and the potential inside the thorax.
the combination of IPC and EIT problems using the same discrete model of the
thorax used to build the inversion matrix from a single patient examination
method to reconstruct the internal thorax impedance parameters from current
measurements and using the boundary element method
simplified impedance image reconstruction algorithm based on the NewtonKantorovich method and on the inversion of non-rectangular tall matrix (notion
introduced by the author)
method and instrument to non-invasive determination of the activation sequence of
the human heart based on the activation front concept
hardware system used to measure the thorax potential evolution
software to pre-process the measured signals
10
Ph.D. Thesis
Extended abstract
software to describe and model the real patient thorax based on geometrical
measurements
software to display the acquired and reconstructed signals, the potential maps, the
animated evolution of potential field and the 3-dimensional activation maps
Selected References
[1] H.Geertjan and Adriaan van Oosterom, The depolarisation sequence of the human heart
surface computed from measured body potentials, IEEE TBME, pp. 1047-1058, Dec 1988.
[2] W.K.Kim., W.Desk, L.E.Bakar and J.Pearce, Origins of the impedance change in impedance
cardiography by a tree-dimensional finite element model, IEEE TBME, pp. 993-1000, Dec
1988.
[3] Adrianus Antonius Hubertus Damen, On the observability of electrical cardiac sources,
Ph.D. Dissertation, Technische Hogeschool Eindhoven.
[4] M.Shenasa, M.Borggrefe and G.Breithart, Cardiac Mapping, Mount Kisco, NY: Futura, 1993.
[5] Gadd R., P.M.Record and P.Rolfe: Finite element modelling for electrical impedance
tomography - AIC of BME vol-12 pag.133-134 / 1990;
[6] Murai T. and Y.Kagawa: Electrical Impedance Computed Tomography Based On A Finite
Element Model - TBME vol BME-32, pag. 177-184 / 1985;
[7] Cristescu A., Stefan POPESCU, K.Schob, and M.Hancu, An integrated system for body
surface heartpotential distribution maps, in Timisoara medicala Tom 32/1987 Supplement,
pp.21-27;
[8] S.Pasca and Stefan POPESCU, Sistem de achizitie de date si sontrol pentru aplicatii
biomedicale - in volumul REP, Snagov 6-7 IX 1990, pag.V20-V25
11