Академический Документы
Профессиональный Документы
Культура Документы
Abstract—Synthesis of the 12-lead ECG has been investigated situations, recording a standard 12-lead ECG, the only ECG
in the past decade as a method to improve patient monitoring representation that cardiologists have been trained to accurately
in situations where the acquisition of the 12-lead ECG is cum- analyze and interpret, is often difficult and impractical. It would
bersome and time consuming. This paper presents and assesses a
novel approach for deriving 12-lead ECGs from a pseudoorthogo- thus be valuable to design a minimal, easy to use lead set from
nal three-lead subset via generic and patient-specific nonlinear re- which the 12-lead ECG can be accurately reconstructed [4].
construction methods based on the use of artificial neural-networks The solution adopted by the enhanced personal, intelligent
(ANNs) committees. We train and test the ANN on a set of serial and mobile system for early detection and interpretation of cardi-
ECGs from 120 cardiac inpatients from the intensive care unit ological syndromes (EPI-MEDICS) European project was to de-
of the Cardiology Hospital of Lyon. We then assess the similar-
ity between the synthesized ECGs and the original ECGs at the sign an intelligent, portable Personal ECG Monitor (PEM), em-
quantitative level in comparison with generic and patient-specific bedding advanced decision making, which is capable to record
multiple-regression-based methods. The ANN achieved accurate a simplified but professional quality three-lead ECG, to derive
reconstruction of the 12-lead ECGs of the study population using a standard 12-lead ECG to detect arrhythmias and ischemia or
both generic and patient-specific ANN transforms, showing sig- acute infarction, and to send an alarm message with a copy of
nificant improvements over generic (p-value ≤ 0.05) and patient-
specific (p-value ≤ 0.01) multiple-linear-regression-based models. the electronic health record (EHR) and the concerned ECG to
Consequently, our neural-network-based approach has proven to the appropriate health care providers [3].
be sufficiently accurate to be deployed in home care as well as in To achieve this goal, the EPI-MEDICS project determined an
ambulatory situations to synthesize a standard 12-lead ECG from easy-to-use pseudoorthogonal set of ECG leads, based on four
a reduced lead-set ECG recording. active electrodes located in the Mason Likar positions and in V 2,
Index Terms—Acute ischemia, arrhythmia, ECG, eHealth, respectively, which have proven to be adequate for home care
embedded computing, neural networks, self-care. or ambulatory ECG recording [5]. The project also determined
a generic linear transformation matrix used to derive the five
I. INTRODUCTION missing leads (V 1, V 3, V 4, V 5, and V 6) of the standard 12-
lead ECG from the recorded three-lead PEM ECG. The overall
ECENT years have witnessed a growing interest for devel-
R oping personalized and nonhospital-based care systems to
improve the management of cardiac care, to reduce the time be-
usability and diagnostic accuracy of the PEM device have been
judged very high. However, it is known that the accuracy of the
standard 12-lead ECG reconstruction can be further enhanced
fore treatment, and consequently, to reduce cardiac morbidity by using patient-specific transforms that could be stored in the
and mortality [1]–[3]. The reason behind such interest is due smart media card embedded in the PEM device. The challenge is
to the fact that in western countries, cardiovascular diseases are thus to design a synthesis method that is able to derive a 12-lead
a leading cause of early disability and premature death. Fur- ECG from the pseudoorthogonal three-lead ECG that contains
thermore, because the population is rapidly aging, the number as much diagnostic information as would have been contained
of cardiac patients is steadily increasing and almost 2/3 of the in the original 12-lead ECG if recorded simultaneously with the
cardiac patients die in the prehospital phase. Therefore, greater three-lead subset.
deployment of resources for prehospital care is needed to reduce Generally, methods used to derive the reconstruction trans-
the fatality rate. forms are based on multiple-linear-regression techniques
The only easy to use diagnosis tool useful for assessing the [6]–[8]. However, most of the studies using linear patient-
probability of cardiac events in home care, self-care, ambulatory, specific transforms have only been assessed on subsets of the
or emergency recording conditions is the ECG. However, in such standard 12-lead ECG recordings where the training and testing
phases were performed on the same data or, in the best cases,
the validation was conducted over ECGs recorded within laps
Manuscript received November 3, 2006; revised March 15, 2007 and May 30, of time within 24 h with the same electrode positions as for the
2007. First published April 8, 2010; current version published June 3, 2010. learning phase. Also, it is essential to evaluate the accuracy of
The authors are with the Department of Methodologies of Information Pro-
cessing in Cardiology, Institut National des Sciences Appliquées (INSA-Lyon),
lead reconstruction over ECGs acquired using personalized and
Institut National de la Santé et de la Recherche Médicale (INSERM), Uni- nonhospital-based care systems [9], such as the PEM system.
versité Lyon 1, Bron, F-69677, France (e-mail: hussein_atoui@hotmail.com; In addition, even though linear transforms yield high accu-
jocelyne.fayn@insa-lyon.fr; paul.rubel@insa-lyon.fr).
Digital Object Identifier 10.1109/TITB.2010.2047754
racy when electrode positions remain in the same place [6]–[8],
II. MATERIAL AND METHODS Fig. 1. Architecture of the ANNs used to synthesize the five missing V 1, V 3,
V 4, V 5, and V 6 leads (output layer) of the 12-lead ECG using a three-lead
A. Study Population ECG (I, II, and V 2) as input layer, h (typically 1 or 2) hidden layers and n
neurons per hidden layer.
We used a database of 10 s ECGs that have been recorded on
cardiac inpatients from the intensive care unit of the Cardiology
Hospital of Lyon. Each patient had one or two pairs of ECGs,
B. Neural-Network Architecture Design and Training
each consisting of the following:
1) a standard 10 s 12-lead ECG, hereafter called reference To synthesize the missing (V 1, V 3, V 4, V 5, and V 6) ECG
ECG, recorded by using a standard digital electrocardio- signals from the recorded (I, II, and V 2) 12-lead ECG sub-
graph (Cardiette); set, we use an ensemble of N multilayer feedforward ANNs
2) a 10 s three-lead PEM ECG based on I, II, and V 2 (PECG). trained by means of a supervised back-propagation algorithm.
The PEM ECGs were recorded in average at less than half an Each individual ANN consists of one input layer with three
hour interval of the recording of the reference ECGs. input neurons (one for each recorded signal), one output layer
An extensive work of quality assessment was performed on with five output neurons (one for each derived signal), h = 1
all ECGs of the database, in collaboration with experienced hidden layer and n = 15 neurons per hidden layer, as shown
cardiologists of the hospital. The target was to remove ECGs in Fig. 1. The relations between the input and output layers are
where situations of electrodes inversions, misplacements, or expressed through the weight and biases of the neurons of the
diagnostic changes were noticed. hidden layer. These weight and biases are chosen randomly at
The study population thus consists of a series of 157 pairs of the beginning of the training phase and are then iteratively opti-
digital ECGs (sampling rate: 500 samples/s; resolution: 5 µV) mized during the back-propagation process. The hidden layers
collected on 120 patients (82 male, 38 female; mean age ± SD number, the activation function type, and the numbers of neu-
= 61 ± 15 years). Acute coronary syndrome (ACS) was diag- rons per layer were chosen on the basis of previous experiments
nosed in 46 (29.3%) of the 157 standard 12-lead ECG tracings. and preliminary results obtained in synthesizing biomedical sig-
The ECGs were processed by the Lyon program [14] that ex- nals by nonlinear methods [15]. As usual in approximation
tracted a 1-s-duration representative cycle from every original tasks, a linear activation function is used for the output neu-
10 s ECG record and determined the corresponding P, QRS, and rons. A sigmoid transfer function was chosen for the hidden
T onsets and offsets. Finally, the ECG signal from the represen- layer.
tative cycle was used in the interval [P-onset − 18 ms, T-offset To overcome the performance limitations inherent to the train-
+ 38 ms] for computing the reconstruction transforms. ing process, such as the random selection of the initial settings
The study population was divided into two subsets DS1 and of the ANN weights and biases of the individual networks, we
DS2 of 83 and 74 pairs of ECGs, respectively. Dataset DS1 adopted a solution consisting of building up ANN committees
consists of the 83 patients (53 male and 30 female) of the of N = 50 individual ANN of the Fig. 1 type. The five outputs
study population who had only one pair of ECGs (21 ACS of the ANN committees are obtained by summing up and di-
ECGs and 62 control ECGs). Dataset DS2 consists of the re- viding by N , the outputs of the N individual ANNs. It is thus
maining 37 patients who had two pairs recorded in average at expected that, as an application of the central limit theorem, the
three (range: 1–21) days interval (25 ACS ECGs and 49 control performance of the ANN committee surpasses each individual
ECGs). ANN performance [3], [11].
ATOUI et al.: NOVEL NEURAL-NETWORK MODEL FOR DERIVING STANDARD 12-LEAD ECGs 885
Fig. 2. Reconstruction of a standard 12-lead ECG out of the three-lead PEM ECG. The synthesis of the five missing V 1, V 3, V 4, V 5, and V 6 leads is obtained
by averaging the outputs of a committee of 50 neural networks.
1) Generic Neural-Network Training: To train each of the be determined by ECG-S, we used a fixed number of itera-
N = 50 individual, generic ANN of the committee, we adopted tions K2 that was determined in a similar fashion, as shown in
a cross-validation-like strategy [16] as follows. Section II-B.1, but on a different, independent serial database
The 83 standard 12-lead ECGs of DS1 were randomly divided previously described in [15]. The learning process was then per-
into m = 5 subsets. Each individual network of the committee formed by training each of the N individual ANN on the first
was trained m times, each time leaving out one of the train- reference ECG (ECG-L) of each of the 37 patients of DS2.
ing subsets. The ECGs of the m − 1 subsets were juxtaposed 3) Neural-Network Model Reutilization: Once trained, the
to form one continuous ECG recording. Stopping the learning generic and patient-specific ANN transforms are tested on a
process was performed by monitoring the reconstruction error different dataset than the one used to train the generic and
(rms) on the remaining subset, where the ECGs were also jux- patient-specific committees. This test is intended to assess the
taposed to form one continuous ECG recording. This procedure ANN models performance by simulating their use in real-life
was repeated for each of the N different networks biases and situations and measuring the reconstruction accuracy of both
weights initializations and for each of the m subsets. A fixed transformation models. The simulation process however sub-
number of iterations K1 was then determined by recording, for stantially differs from one reconstruction method to the other.
each training case, the number of iterations k and by averaging In the generic approach, we use the second standard 12-lead
these values over the N × m trainings. Finally, each individ- and PEM ECG pairs of dataset DS2, i.e., ECG-T and the corre-
ual network of the committee was trained over the 83 12-lead sponding PEM ECG from DS2, as a test set. The five “missing”
ECGs of DS1 using the fixed number of iterations K1 as an leads of every standard 12-lead and PEM ECG are synthesized,
early stopping point. as shown in Fig. 2, by the generic ANN committee designed in
2) Patient-Specific Neural-Network Training: Ideally, we Section II-B.1, and the reconstructed V 1, V 3 to V 6 leads are
would need three serial 12-lead ECGs per patient for deriving then compared with the original leads of the corresponding stan-
a patient-specific ANN ensemble: one for training the network dard 12-lead ECG of DS2. In the patient-specific approach, a
(ECG-L), one for stopping the learning process in order to avoid patient-specific committee is trained on the first standard 12-lead
network overdesign (ECG-S), and one for testing the general- ECG (ECG-L) of the first ECG pair of each of the 37 patients of
ization capacity of the trained network (ECG-T). DS2, according to the method described in Section II-B.2. The
Unfortunately, our database comprised at maximum only two patient-specific committee is then tested on each second ECG
serial standard 12-lead ECGs per patient (ECGs L and T). To pair of the corresponding patient by deriving the five “missing”
overcome this difficulty, instead of stopping the learning pro- leads for both the Cardiette (ECG-T) and the PEM ECGs. Fi-
cess after a variable number of iterations that should ideally nally, the reconstructed V 1, V 3 to V 6 leads are compared with
886 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 3, MAY 2010
C. Multiple-Regression-Synthesis-Coefficients Calculation
To calculate a generic multiple regression transform, we con-
sidered the 83 standard 12-lead ECGs of dataset DS1. As for
the generic ANN training, these 83 ECGs were juxtaposed into
one continuous ECG recording. We then computed a global
transformation matrix that synthesizes the missing V 1, V 3 to
V 6 leads of the PEM ECG, using a multiple-linear-regression
algorithm of the type
Vi = ai0 + ai1 I + ai2 II + ai3 V 2. (1)
(PECG-T) associated with the second Cardiette ECG (ECG-T).
Patient-specific multiple-regression-based transforms were Then, we conducted another correlation analysis between leads
obtained by using the same algorithm to compute an individual I, II, and V 2 of the second Cardiette ECG (ECG-T) and the
transformation matrix for every first 12-lead ECG (ECG-L) of second PEM ECG (PECG-T) of dataset DS2.
every patient of dataset DS2.
D. Validation Database (DS2) Three-Lead and 12-Lead ECG E. Generalization Capability Verification and Validation
Signals Alignment and Coherence Control To assess the quality of the reconstruction of the generic and
Since we were using two different recording systems (PEM patient-specific ANN-based transforms, we measured the lead-
and Cardiette) with different electrodes positioning and because by-lead signal differences between the derived and the original
the recording of the two ECGs was not performed at the same ECGs. The overall waveform similarity between the original
time, it was essential to first proceed to the Cardiette and PEM and the reconstructed ECGs was assessed as follows.
ECG signals alignment and to control the coherence between the 1) Low-pass filtering of the original and reconstructed ECG
two serial recordings before measuring the goodness of fit of the signals by means of a 21-term, 40 Hz cutpoint, moving
reconstructed leads of the dataset used for the validation phase: average filter.
DS2. This step is thus prefiguring a real situation of routine use 2) Calculation of the rms and the correlation coefficients be-
of the PEM. tween the QT signal segments of the original and the
1) PEM and Cardiette Signals Alignment: To establish the reconstructed ECG leads of DS2 for both the generic
time shift (delay) between the PEM and Cardiette ECGs, we and patient-specific ANN and multiple-linear-regression
computed the cross correlation between the V 2 lead signals. In models.
order to be less sensitive to artifacts, we only considered the 3) Exclusion for each of the four reconstruction methods,
signal in the following time interval [Q-onset − 18 ms, T-offset and for every analyzed lead of 5% (two cases) of the rms
+ 38 ms]. The algorithm consists of the following steps. and correlation measurements, which have, respectively,
1) Calculation of the cross-correlation between the two the highest rms or the lowest correlation coefficient be-
signals. tween the original and the reconstructed ECGs. This step
2) Estimation of the time shift between both signals by lo- is intended to overcome any bias generated by outliers in
calization of the maximum value in the cross-correlation. the calculation of the mean and standard deviation [17].
3) Alignment of the global QRS onset of the PEM ECG Statistical analysis of the differences between the rms of the
with the global QRS onset of the standard 12-lead ECG ANN and linear-regression-based reconstruction methods was
by subtraction of the time shift measured by the cross- performed by means of a paired t-test and a trimmed–paired
correlation technique. t-test. The trimmed t-test is used to overcome the t-test sensi-
2) PEM and Cardiette Recordings Coherence Control: Be- tivity to outliers. Both tests are performed with the BioMeDical
cause the PEM and the standard 12-lead ECGs have not been statistical software package (BMDP).
recorded synchronously and because of the fast changing car-
diac status usually found in an intensive care unit, there might III. RESULTS
be large differences between the two ECGs. The objective of
this step is to evaluate if the difference between the recorded A. Original ECG-Recordings Coherence Assessment
PEM and Cardiette ECGs is not too large, and thus, could af- Table I displays the rms values and the correlation coefficients
fect the synthesis quality. The coherence control is performed between the first and second Cardiette ECGs of DS2, after re-
in two steps: for each of the 37 patients of DS2, we conducted moval of the 5% extreme values (N = 35). The average rms
a correlation analysis between leads I, II, V 1, V 2 to V 6 of the signal difference between the standard I, II, V 1, V 2, V 3, V 4,
first and second Cardiette ECGs and between leads I, II, and V 2 V 5, and V 6 leads in the time interval [Q-onset − 18 ms, T-offset
of the first Cardiette ECG (ECG-L) and the second PEM ECG + 38 ms] is 107 µV. The average correlation coefficient is 0.92.
ATOUI et al.: NOVEL NEURAL-NETWORK MODEL FOR DERIVING STANDARD 12-LEAD ECGs 887
TABLE II TABLE IV
RMS (IN MICROVOLTS) AND CORRELATION COEFFICIENTS (r) BETWEEN THE RMS (IN MICROVOLTS) AND CORRELATION COEFFICIENTS (r) BETWEEN THE
ORIGINAL I, II, AND V 2 LEADS OF THE FIRST CARDIETTE ECG AND THE ORIGINAL AND THE RECONSTRUCTED V 1, V 3 TO V 6 LEADS FOR THE
SECOND PEM ECG RECORDINGS FOR DATASET DS2 (N = 35) GENERIC ANN COMMITTEE (ANN-G) AND REGRESSION-BASED
MODELS (REG-G) FOR DATASET DS2 (N = 35)
TABLE III
RMS (IN MICROVOLTS) AND CORRELATION COEFFICIENTS (r) BETWEEN THE
ORIGINAL I, II, AND V 2 LEADS OF THE SECOND CARDIETTE ECG AND THE
CORRESPONDING PEM ECG RECORDINGS FOR DATASET DS2 (N = 35)
TABLE V
RMS (IN MICROVOLTS) AND CORRELATION COEFFICIENTS (r) BETWEEN THE
ORIGINAL AND THE RECONSTRUCTED V 1, V 3 TO V 6 LEADS FOR THE
PATIENT-SPECIFIC ANN COMMITTEE (ANN-S) AND REGRESSION-BASED
MODELS (REG-S) FOR DATASET DS2 (N = 35)
IV. DISCUSSION
In this paper, we designed and assessed the efficacy of
a nonlinear neural-network approach for missing ECG leads
synthesis.
B. Generic Versus Patient-Specific Reconstruction home healthcare, and disease prevention,” IEEE Trans. Inf. Technol.
Biomed., vol. 9, no. 3, pp. 325–336, Sep. 2005.
Patient-specific transforms have proven to be more accurate. [2] P. Campbell, J. Patterson, D. Cromer, K. Wall, G. L. Adams, A. Albano,
Nevertheless, generic reconstructions remain particularly attrac- C. Corey, P. Fox, J. Gardner, B. Hawthorne, J. Lipton, M. Sejersten,
A. Thompson, A. Thompson, S. Wilfong, C. Maynard, and G. Wagner,
tive in situations, where no standard 12-lead ECG could be “Prehospital triage of acute myocardial infarction: Wireless transmission
recorded and/or no patient-specific transform could be com- of electrocardiograms to the on-call cardiologist via a handheld computer,”
puted and/or retrieved, prior to the recording of a reduced ECG J. Electrocardiol., vol. 38, no. 4, pp. 300–309, 2005.
[3] P. Rubel, J. Fayn, L. Simon-Chautemps, H. Atoui, M. Ohlsson, D. Telisson,
lead set. Also, generic reconstructions are less sensitive to arti- S. Adami, S. Arod, M. C. Forlini, C. Malossi, J. Placide, G. L. Ziliani,
facts and electrode displacements, since their coefficients were D. Assanelli, and P. Chevalier, “New paradigms in telemedicine: Ambient
computed from a very large set of ECGs of different origins. intelligence, wearable, pervasive and personalized,” Stud. Health Technol.
Inf., vol. 108, pp. 123–132, 2004.
For example, during the EPI-MEDICS project evaluation, we [4] B. J. Drew, M. M. Pelter, S. F. Wung, M. G. Adams, C. Taylor, G. T. Evans
found a few PEM ECG tracings in which lead V 2 could dis- Jr, and E. Foster, “Accuracy of the EASI 12-lead electrocardiogram com-
play up to 50% changes of the QRS peak-to-peak amplitude pared to the standard 12-lead electrocardiogram for diagnosing multiple
cardiac abnormalities,” J. Electrocardiol., vol. 32, pp. 38–47, 1999.
between two recordings performed at a few days interval. In [5] P. Rubel, J. Fayn, G. Nollo, D. Assanelli, B. Li, L. Restier, S. Adami,
these cases, the generic transform performed better because the S. Arod, H. Atoui, M. Ohlsson, L. Simon-Chautemps, D. Télisson, C.
generic reconstruction of V 1, V 3, and V 4 was less influenced Malossi, G. L. Ziliani, A. Galassi, L. Edenbrandt, and P. Chevalier, “To-
ward personal eHealth in cardiology. Results from the EPI-MEDICS
by the amplitude of V 2 than was the patient-specific transform. telemedicine project,” J. Electrocardiol., vol. 38, no. 4 (suppl.), pp. 100–
To optimize the ECG synthesis process, it would thus be valu- 106, 2005.
able to implement an algorithm that first checks the quality of [6] S. P. Nelwan, “Evaluation of 12-lead ECG reconstruction methods for
patient monitoring,” Ph.D. dissertation, Erasmus Universiteit Rotterdam,
the recorded ECG signals for artifacts or significant amplitude Rotterdam, The Netherlands, 2005.
changes before applying the patient-specific transform and, in [7] J. A. Scherer, J. M. Jenkins, and J. M. Nicklas, “Synthesis of the 12-Lead
case of large differences, either request the user, whenever pos- electrocardiogram from a 3-lead subset using patient-specific transforma-
tion vectors. An algorithmic approach to computerized signal synthesis,”
sible, to check the electrode positions, or switch to the less J. Electrocardiol., vol. 22, pp. 128–136, 1989.
sensitive generic 12-lead ECG synthesis method. [8] S. P. Nelwan, J. A. Kors, S. H. Meij, J. H. van Bemmel, and M. L. Simoons,
“Reconstruction of the 12-lead ECG using reduced lead sets for patient
monitoring,” in Einthoven 2002: 100 years of Electrocardiography. M.
C. Study Limitations J. Schalij, M. J. Janse, A. van Oosterom, H. J. J. Wellens, E. E. van der
Wall, Eds. Leiden, the Netherlands: The Einthoven Foundation, 2002, pp.
In this study, we could only partly assess the performance of 545–551.
the patient-specific ANN synthesis method, since we lack a third [9] L. Hadzievski, B. Bojović, V. Vukcević, P. Belicev, S. Pavlović, Z. Vasil-
serial ECG recording for stopping the learning process, and we jević-Pokrajcić, and M. Ostojić, “A novel mobile transtelephonic system
with synthesized 12-lead ECG,” IEEE Trans. Inf. Technol. Biomed., vol. 8,
had to use a fixed number of iterations to avoid overfitting that no. 4, pp. 428–438, Dec. 2004.
could limit the generalization capabilities. [10] R. P. W. Duin, “Learned from neural networks,” in Proc. 6th Annu. Conf.
Adv. School Comput. Imag. (ASCI 2000), pp. 9–13.
[11] S. Mitra, S. K. Pal, and P. Mitra, “Data mining in soft computing frame-
V. CONCLUSION work: A survey,” IEEE Trans. Neural Netw., vol. 13, no. 1, pp. 3–14, Jan.
2001.
This paper investigates the accuracy of the reconstruction of [12] R. Modre, M. Seger, G. Fischer, C. Hintermuller, D. Hayn, B. Pfeifer,
the 12-lead ECG from a reduced lead set using generic and F. Hanser, G. Schreier, and B. Tilg, “Cardiac anisotropy: Is it negligible
regarding noninvasive activation time imaging?” IEEE Trans. Biomed.
patient-specific ANNs. The results suggest that ANN represent Eng., vol. 53, no. 4, pp. 569–580, Apr. 2006.
a rather interesting and very promising approach to improve the [13] R. Gulrajani, “The forward and inverse problems of electrocardiography,”
current 12-lead ECG synthesis method based on linear trans- IEEE Eng. Med. Biol. Mag., vol. 17, no. 5, pp. 84–101, Sep./Oct. 1998.
[14] P. Arnaud, P. Rubel, D. Morlet, J. Fayn, and M. C. Forlini, “Methodology
forms, especially for the patient-specific approach. The differ- of ECG interpretation in the lyon program,” Methods Inf. Med., vol. 29,
ences between the original and the reconstructed ECGs were no. 4, pp. 393–402, 1990.
mainly ascribable to electrode displacements or to hour-to-hour [15] H. Atoui, J. Fayn, and P. Rubel, “A neural network approach for patient-
specific 12-lead ECG synthesis in patient monitoring environments,” in
changes in the ECG and not to the synthesis method. ECG cases Computers in Cardiology, vol. 31, A. Murray, Ed., 2004, pp. 161–164.
in which the rms were high or the correlations were low were an- [16] B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap. London,
alyzed by an experienced cardiologist. His conclusion was that U.K.: Chapman & Hall, 1993.
[17] “Recommendations for measurement standards in quantitative electrocar-
the differences between synthesized and original leads are diag- diography. The CSE Working Party,” Eur. Heart J., vol. 6, pp. 815–825,
nostically irrelevant in the majority of cases. The patient-specific 1985.
transform could be further improved if two serial standard 12-
lead ECGs are available for the learning process. We also believe
Hussein Atoui received the B.S. degree in computer
that the generic ANN synthesis method can be further optimized science from Lebanese University, Beirut, Lebanon,
by using a larger and more representative training database than in 2001, and the M.S. and Ph.D. degrees in in-
the one used in this study and should clearly supersede the linear formatics from the Institut National des Sciences
Appliquées de Lyon (INSA-Lyon), Lyon, France, in
transform. 2002 and 2006, respectively.
From 2002 to 2008, he has been with the
Méthodologies de Traitement de l’Information en
REFERENCES Cardiologie (MTIC) Research Laboratory of INSA-
[1] F. Axisa, P. M. Schmitt, C. Gehin, G. Delhomme, E. McAdams, and Lyon and Université Lyon 1, Lyon. He is currently Re-
A. Dittmar, “Flexible technologies and smart clothing for citizen medicine, search Engineer in an IT solutions provider company.
890 IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 3, MAY 2010
Jocelyne Fayn received the M.S. degree in mathe- Paul Rubel (M’91) received the degree in electron-
matics, specialty statistics and informatics, from the ics engineering from the Institut National des Sci-
Université Claude Bernard Lyon 1, Lyon, France, ences Appliquées (INSA), Lyon, France, in 1966,
and the Ph.D. degree in computer sciences from the and the Ph.D. degree in electronics from the Univer-
Institut National des Sciences Appliquées (INSA- sité Claude Bernard Lyon 1, Lyon, in 1970.
Lyon), Lyon. She is currently Research Engineer in He is currently a Professor of informatics in the
the Institut National de la Santé et de la Recherche Department of Informatics, INSA de Lyon, and has
Médicale (INSERM) and Assistant-Director of the been an Advisor of circa 20 Ph.D. theses in medical
Méthodologies de Traitement de l’Information en informatics at INSA and the University of Lyon 1.
Cardiologie (MTIC) Research Laboratory of INSA- From 1986 to 2000, he was the Head of the Research
Lyon and Université Lyon 1, Lyon. She is also the Unit U121 on the “Electrical Activity of the Heart”
Coordinator of the study unit Medical Informatics and Communication Tech- of the National Institute of Health and Medical Research (INSERM), and he is
nologies of the Master Biomedical Research, University of Lyon. She was en- currently the Head of research team MTIC on “Information Processing Method-
gaged in three national French projects and eight UE-funded research projects. ologies in Cardiology” of INSA-Lyon and Université Lyon 1, Lyon. He has peer
She is an International Editor of the Journal of Electrocardiology and of the reviewed several international journals and conferences. He has authored or
International Journal of Telemedicine and Applications. She was a Reviewer of coauthored more than 220 publications.
a number of international journals. Her research interests include information Prof. Rubel was the Project Coordinator of the European Advanced Informat-
systems, pervasive computing, ambient intelligence, information and communi- ics in Medicine and Information Society Technology projects: Open European
cation technologies, medical informatics, quantitative electrocardiology, serial Data Interchange and Processing for Electrocardiography and Enhanced Per-
ECG analysis, and medical decision-making. She has authored or coauthored sonal, Intelligent, and Mobile system for Early Detection and Interpretation of
more than 100 publications. Cardiac Syndromes, and was the local Project Manager of AIM Project standard
Dr. Fayn was the Chair of the 32nd IEEE Annual Conference on Com- communications protocol-computer-assisted electrocardiography (SCP-ECG),
puters in Cardiology. She was an Invited Expert of European Committee for and of five additional AIM and Health Telematics Applications Projects. He has
Standardization (CEN)/Technical Committee 251/Working Group (WG) 4 on also been a member of the Steering Committee of the European concerted ac-
Health Informatics Technology for interoperability and a member of the board tion “Common Standards for quantitative Electrocardiography” (CSE) (DGXII,
of the WG15 working group Computers in Cardiology of the European Society 1978–1990) for several years and is currently responsible for the world-wide
of Cardiology for several years. She is a member of program committees of distribution of the CSE databases and for the quality assessment of ECG in-
several international conferences and workshops. terpretation programs since 1995. He has also been an active Expert of Euro-
pean Committee for Standardization (CEN)/Technical Committee 251/Working
Group 4 on Health Informatics Technology for interoperability, and was one of
the three Core-Team experts, which were responsible for the development of
the so called SCP-ECG CEN EN 1064 European Norm.