Olena Punshchykova 1,3; Peter Kneppo 1; Milan Tyšler 1,2
Faculty of Biomedical Engineering, Czech Technical University in Prague, Kladno, Czech Republic
1; Institute of Measurement Science, Slovak Academy of Science, Bratislava, Slovakia
2; Intercollegiate Faculty of Medical Engineering, National Technical University of Ukraine
“Kyiv Polytechnic Institute”, Kyiv, Ukraine
Lékař a technika - Clinician and Technology No. 2, 2012, 42, 7-10
Conference YBERC 2012
Novel noninvasive method for localization of areas of myocardium with changed repolarization that can lead to life threatening cardiac arrhythmias is presented. Method is based on multichannel ECG measurement, body surface potential mapping and solving the inverse problem of electrocardiology to find one or two dipoles representing the pathological sources in the heart. Body surface potential mapping system ProCardio-8 using up to 144 active electrodes was used for multichannel ECG measurement. For the inverse calculations, inhomogeneous model of the shape of thorax with lungs and ventricles including cavities was used. Ability of the system to identify selected cardiac pathologies is demonstrated on cases with one or two ischemic foci. In this paper, analysis of existing system and algorithms for cardiac pathology localization is given in detail. Further software development will include ventricular premature beat foci localization.
Keywords: multichannel ECG recording, body surface potentials, inverse problem of electrocardiography, repolarization changes, arrhythmia
According to the European Heart Rhythm
Association White Book 2011, in 2009 the proportion
of deaths resulting from cardiovascular diseases was
50% in the Czech Republic and 65% in Ukraine .
Therefore it is vital to decrease these numbers using
new diagnostic and treatment methods.
Necessary precondition for the use of advanced
diagnostic methods enabling precise cardiac pathology
localization in various cardiovascular diseases is
availability of quality high resolution measurement of
Diagnostic information contained in body surface
potential maps and acquired from multichannel ECG
signals is much more detailed than that obtained from
standard 12-lead ECG or Frank VCG lead system.
For high quality measurement of required ECG
signals special multichannel electrocardiographic
system ProCardio-8 was developed .
Material and Methods
Body surface potential mapping (BSPM) is a
technologically advanced method that helps in detailed
diagnostics of various cardiovascular diseases. Using
BSPM for localization of areas with changed
repolarization can significantly improve the
understanding of electrophysiological processes in the
heart in different clinical cases. The procedure is
noninvasive and non-fluoroscopic and can provide
highly useful diagnostic information without
substantially loading the patient. In addition, its safety,
possibility of repeated examinations and their
automation gives BSPM more important advantages in
comparison to other diagnostic procedures.
BSPM is based on multichannel measurement of
body surface cardiac potentials using special
multichannel electrocardiograph. Detailed registration
of the surface cardiac potentials requires high number
of sensing electrodes accurately placed on the thorax.
BSPM is one of the several cardiac mapping
methods. “Cardiac mapping” is a wide term which also covers endocardial and epicardial mapping.
Endocardial mapping is performed with catheters,
which are inserted into heart cavities and navigated
using fluoroscopy. Inserted catheters sequentially
record endocardial electrograms and enable to identify
temporal and spatial distributions of electrical
potentials generated by the myocardium during normal
and abnormal rhythms.
In contrast to the above mentioned two methods of
cardiac mapping, noninvasive assessment of the
electrical state of the heart by solution of the inverse
problem of electrocardiography requires high quality
surface ECG measurement with correct and precise
electrode positioning and an appropriate model of the
patient torso volume conductor. This model can be
either some simplified approximation of a real torso, or
can be precisely derived from CT or MRI scans of the
patient torso what considerably increases accuracy of
the method. The variety of cardiovascular pathologies
that can be identified using such noninvasive inverse
procedure includes ischemia, WPW, ventricular ectopy
and other arrhythmias .
Multichannel ECG mapping system ProCardio-8
In order to record and analyse high quality
multichannel ECG recordings, a special system
ProCardio-8 was designed. This device consists of a set
of active electrodes (up to 144), a data acquisition
system and a hosting personal computer. Rechargeable
Li-ion battery power supply with advanced power
management is used for achieving excellent recordings
quality and possibility of its using for one working day
without replacement or recharging.
The data acquisition system is placed in a patient
terminal box and connected via USB cable to the host
personal computer. Its small geometric dimensions
(14 × 19 × 20 cm) minimize capacitive coupling with
external environment. The data acquisition system is
modular and built from up to 9 measuring boards that
are plugged into a motherboard. The system can be
configured to record ECG signals simultaneously from
64 to 144 electrodes. Each measuring board records 16
ECG signals. One of the boards is configured as the
reference board and is used for recording of 3 signals
from the limb electrodes R, L and F. The mean value
from the limb electrodes - the potential of the Wilson's
central terminal (WCT) that is commonly used as the
reference for all unipolar leads - is generated in a
resistor network as the 4th signal from the board.
Resting 12 channels on the board can be used for
signals from additional chest electrodes.
The reference board also contains additional circuitry
for the common mode sense electrode (CMS) and the
driven right leg (DRL) electrode that provides active
grounding of the patient and reduces the common
mode voltage. DRL circuit also limits the current
through the patient body to 50 μA. For patient
protection in case of possible electrical defects,
additional protection circuit is used that generates a
power-down signal in case when the current through
the patient body remains close to 50 μA. In such a case
the microcontroller disables the power supply of the
To get optimal signal quality, all recorded signals are
originally measured relatively to the properly placed
CMS electrode. Potentials of unipolar
electrocardiographic leads are then computed in the PC
by subtracting the WCT potential from the electrode
signals, while bipolar leads are computed as difference
signals between corresponding electrodes. Each
measuring channel is low-noise (< 0.75 μVRMS) and
equipped with a DC-coupled instrumentation amplifier
(Analog Devices AD 627) with fixed gain of 40 and a
22-bit Σ-δ A/D converter (Analog Devices AD 7716).
Sampling frequency can be selected between 125 and
2000 Hz and provides corresponding effective dynamic
resolution between 19 and 16 bits.
The data acquisition system is controlled by a 16-bit
CISC Fujitsu microcontroller placed on the
motherboard. It scans the data sampled from analogue
channels and sends it to the connected host PC. To
minimize the capacitive coupling between the USB
port of the host PC and the patient terminal, a fibre
optic USB extension cable is used.
Disposable Ag-AgCl electrodes with active adapters
are used for ECG signals sensing. Each electrode is
made in SMD technology and has a thermally
compensated amplifier (Analog Devices OP 193).
Therefore each electrode has high input and very low
output impedance, what effectively reduces disturbing
signals, guarantees low noise induced in electrode
cables and enables high quality ECG measurement.
Disposable electrodes usage eliminates risk of patient
infections as well.
The host personal computer of the ProCardio-8
system runs Microsoft Windows® based measuring and
data analysis application software containing five basic
Configuration of the data acquisition system,
Testing of electrode contacts and signal calibration,
Measurement of ECG signals,
Off-line ECG signal processing,
Computation and analysis of body surface potential maps.
The first three user-friendly programmed modules
enable to set all necessary parameters of the system, to
test electrode-skin connections and optimize signal
processing depending on ECG amplitudes, read the
ECG data stream into the host PC, monitor recorded
ECG signals on the screen and store recorded
measurements. Measured and processed data are stored
as GDF files (general data format for biosignals) in
selected directory. The fourth module performs off-line
ECG signal filtering and baseline correction, enables to
select desired time instants in the ECG records and
finally the fifth module computes, displays and analyses body surface potential maps, integral maps
and desired difference or departure integral maps.
Creating, debugging and consecutive testing of any
real-time measuring software is a very difficult and
time consuming process, especially in so sensitive field
as medical sciences. The classical approach is to
develop an application source code in C++ language
what certainly takes a large amount of programmer’s
time and labour. To simplify and accelerate this
process, we decided to use the MATLAB environment
and its supporting background also for measurement
control and communication with the external hardware.
This concept offers the opportunity to utilize all digital
signal processing and analysis functions included in the
MATLAB for processing of the data stream during the
data acquisition and subsequent off-line processing and
analysis of measured ECG signals .
Cardiac pathology localization method
The location of some cardiac pathology can be found
noninvasively by solving the inverse problem of
electrocardiology. In our study, possibility to locate
one or two ischemic areas with changed repolarization
The noninvasive inverse method was based on a
dipole model of the cardiac electric generator
representing small pathological lesions and realistic
model of the geometry and electrical properties of an
inhomogeneous human torso , .
The method evaluates differences between surface
potentials recorded in the same patient under
conditions without manifestation of ischemia and
during ischemia developed during a stress test or by
progress of the disease.
Differences within the depolarization - repolarization
period can be topologically represented as one
difference integral map (DIM) showing the surface
distribution of differences between integrals of ECG
potentials over the QRST interval recorded in
corresponding surface points. Assuming a multiple
dipole generator and piecewise homogeneous torso as
the volume conductor, boundary element method can
be applied for computation of body surface potentials
yielding a linear matrix equation.
Computation of the DIM by subtracting the normal
integral map from the integral map with manifested
ischemia is then equivalent to the computation of an
integral map for a difference multiple dipole generator
Δs representing only the pathological changes:
where Δp is the vector of computed values in the DIM,
A is the transfer matrix representing the torso volume
conductor, si and sn are dipolar sources in the normal
and ischemic myocardium and Δs represents dipolar
generators in ischemic lesions; physically s represents
time integrals of dipole moments of current dipoles.
In the inverse solution, a single dipole or a pair of
dipoles was searched as the equivalent generator (EG)
representing the small lesions. The inverse solution
was based on singular value decomposition of a
submatrix of the transfer matrix A. The submatrix was
composed of columns of the matrix A corresponding to
every possible position of an EG located in the
predefined points evenly distributed within the
modelled ventricular myocardium (N possible dipoles,
or N.(N-1)/2 possible dipole pairs).
As the transfer submatrix for each dipole or pair of
dipoles is strongly over-determined, unique solution in
the sense of the minimum least-squares criterion can be
always obtained. The best representative dipole or pair
of dipoles with minimal rms difference (DIF) between
the measured DIM and the map generated by the
inversely estimated equivalent generator was then
considered to represent the analysed cardiac pathology.
However, especially for dipole pairs, this criterion
usually had no sharp minimum and several dipole pairs
gave results with DIF varying very slightly from the
minimum. Therefore also results with DIF within 1%
difference from the best solution were analysed. To
identify the cases, where the inversely computed
dipoles really represented 2 lesions, two clusters of
dipoles were created from the obtained dipole pairs by
applying the modified K-means iterative algorithm for
K=2, based on Euclidean distance between the dipoles.
Dipole positions of the dipole pair with smallest DIF
value were used as the initial positions of the cluster
centres. Dipoles from another dipole pairs were then
assigned to the cluster with the nearest cluster centre.
Because dipoles from one pair should represent
different ischemic lesions, they should belong to
different clusters. If both dipoles were assigned to the
same cluster, the pair was excluded from the
evaluation. At the end of each iteration, new cluster
centres were recalculated from assigned dipoles and
next iteration was started. If no more changes occurred
during the iteration, the algorithm finished dividing
dipoles into 2 clusters. The final gravity centre of each
cluster was considered as the centre of an identified
lesion and the mean dipole moment computed from all
dipoles in one cluster was assigned to that lesion. At
the end of the iteration process, the DIM was claimed
to represent two distinct lesions only if all pairs of the
inverse dipoles have their dipoles located in different
clusters and the distance between the clusters was big
To locate the heart repolarization changes, body
surface potentials from multiple chest leads were
measured using the ProCardio-8 in a group of patients
(8 men and 3 women, age 45-69) after myocardial
infarction and surface QRST integral maps were
computed. In order to calculate equivalent current
dipoles representing the regions with changed
repolarization, inhomogeneous torso model with lungs
and both, realistic and analytical heart model were
used. Integral maps before and during the exercise test
at load of 75W were computed for each patient and the
corresponding EG was considered as the representation
of possible local repolarization changes induced due to
the stress test and was compared with perfusion images
obtained by single photon emission computer
tomoghraphy (SPECT). Analysis of the results showed
that in 80% of cases the EG representing the QRST
DIMs was located in areas identified also by the
SPECT, so the method could be a useful tool for
localization of ischemic heart lesions .
Moreover, information on individual torso structure
from MRI, CT or ultrasound systems could
significantly increase the quality and accuracy of this
advanced inverse diagnostic method for identification
of local repolarization changes .
Ventricular arrhythmias belong to the main causes of
mortality in the western society. Ablation of the ectopic
centre, which causes ventricular cardiac rhythm
pathology, can significantly improve the stability of the
heart function and decrease the number of life
threatening arrhythmias. O. Dössel’s group  has
used anatomical data of the patient and a model based
on the cellular automaton principle to demonstrate
possibility of ectopic focus localization. Their method
showed reliable localization of premature ventricular
beat foci with reconstruction error less than 6.1 mm.
Therefore we are going to supplement the ProCardio-8
software with new algorithms which will enable to
detect ventricular ectopic centres using the Tikhonov
regularization method .
Analysis of the normal electrophysiology of the
heart, cardiac arrhythmia mechanisms origin and ECG
genesis specifics showed that using body surface
potentials together with patient torso model (using
cardiac CT or MRI scans) is possible instrument for
cardiac arrhythmia substrate localization. Physical and
mathematical aspects of the inverse solution using
simplified dipole-based model of cardiac pathologies
were experimentally verified and ProCardio-8 mapping
system, for computation and evaluation of body surface
potential maps was developed. Efficiency of the
solution of the inverse problem of electrocardiography
and its possibility to locate cardiac pathologies was
demonstrated on localization of one or two
simultaneous ischemic lesions with changed
Further software development will include
ventricular premature beat foci localization.
Application of this method for arrhythmia diagnostics
can expand understanding of electrophysiological
processes in the heart in different clinical cases.
This work has been supported by the research grants
No. NT/11532-5 from the Ministry of Health of the
Czech Republic IGA, No. SGS11/143/OHK5/2T/17
from Czech Technical University in Prague SGS, No.
2/0210/10 from the VEGA Grant Agency and No.
APVV-0513-10 from the Slovak Research and
Olena Punshchykova, Mgr.
Department of Biomedical Technology
Faculty of Biomedical Engineering
Czech Technical University in Prague
nám. Sítná 3105, CZ-272 01 Kladno
Phone: +420 777 093 734
 Vardas, P., Auricchio, A., Merino, J.L. The EHRA White Book 2011. The Current Status of Cardiac Electrophysiology in ESC Member Countries. p. 498, www.escardio.org/EHRA.
 Karas, S., Hebláková, E., Rosík, V., Tyšler, M. Multichannel high-resolution electrocardiograph developed in Matlab. Technical Computing Prague. 16th Annual Conference Proceedings, Praha, Czech Republik, 2008, p. 55.
 Tyšler, M., Rosík, V., Kneppo, P. Multichannel ECG measurement for noninvasive identification of heart regions with changed repolarization. XVIII IMEKO World Congress. Metrology for a Sustainable Development, Rio de Janeiro, Brasil, 2006.
 Rosík, V., Karas, S., Hebláková, E., Tyšler, M., Filipová, S. Portable device for high resolution ECG mapping. Measurement science review, 2007, vol. 7, no. 6, p. 57-61.
 Tyšler, M., Svehlikova, J., Rosík, V., Karas, S., Hebláková, E., Kneppo, P., Muzik, J., Kania, M., Zaczek R., Kobylecka, M. Model-based method and instrumentation for noninvasive identification of local ischemic lesions in the heart. World Congress on Medical Physics and Biomedical Engineering. IFMBE Proceedings, Munich, Germany, 2009, p.1403-1406.
 Svehlikova, J., Lenkova, J., Turzova, M., Tysler, M., Kania, M., Maniewski R. Influence of individual torso geometry on inverse solution to 2 dipoles. J Electrocardiol, 2012, vol. 45, no. 1, p. 7- 12.
 Farina, D., Dössel, O. Non-invasive model-based localization of ventricular ectopic centers from multichannel ECG. International Journal of Applied Electromagnetics and Mechanics, 2009, vol. 30, no. 3-4, p. 289-297.
 Denisov, A. M., Zakharov, E. V., Kalinin, A. V., Kalinin, V. V. Numerical solution of the inverse electrocardiography problem with the use of the Tikhonov regularization method. Moscow University Computational Mathematics and Cybernetics, 2008, vol. 32, no. 2, p. 61-68.