Autonomous Cardiac Diagnostic based on Synchronized ECG and

PCG Signal

Z. Bouguila

, A. Moukadem

, A. Dieterlen

, A. Ahmed Benyahia

2,3

, A. Hajjam

, S. Talha

and E. Andres

MIPS Laboratory, University of Haute Alsace, 61 rue Albert Camus, 68093 Mulhouse, France

Irtes, Université de Technologie Belfort Montbéliard, 90010 Belfort, France

Newel, 36 rue Paul Cezanne, 68200 Mulhouse, France

Hôpital Civil, 1 Place de L’hôpital, 67000 Strasbourg, France

1 STAGE OF THE RESEARCH

The MIPS Laboratory (Modelling, Intelligence,

Process and Systems) is an interdisciplinary research

laboratory hosted by the Haute Alsace University. It

is involved in several research projects that deal with

signal processing, software engineering, microscopy

imaging and modeling and identification in

automatic and mechanic.

Since March 2012, the MIPS laboratory is

involved in the telemedicine project E-care

(www.projet-e-care.fr) piloted by the NEWEL

society. This project rallied economics and scientific

community to keep patients in the comfort of their

own home with a higher level of care and all this

with reduction cost.

The E-care project aims to develop best practices

and a platform for awareness raising, knowledge

exchange and policy making in this field. The

project is closely linked to the thesis contribution of

Ali Moukadem (Moukadem, 2011). Indeed, this

thesis fulfilled in MIPS laboratory and HUS

“Hôpitaux Universitaires de Strasbourg” co-financed

by “region Alsace” and ANR ASAP TLOG 06

project, became interested in development of robust

methods for heart sound analysis which can be used

for auto-diagnosis and telemedicine applications.

Diagnosis based PhonoCardioGram (PCG)

signals alone cannot detect all cases of heart

symptoms (Ahlstrom et al., 2008). In this work, we

are recommended for using an EleCtrocardioGram

(ECG) signal besides a PCG signal for heart disease

investigation. The advantage of the proposed system

is that a heart’s diagnosis system based on the ECG

and PCG signals can lead to high performance heart

diagnostics (Ping and Zhigang, 1998).

This thesis project, brings economic and

scientific partners together, is financially supported

by “Caisse des Dépots” and “region Alsace” and

seeks to improve the use of telemedicine for life

saving. The activities planned focus on providing

good practices and improving the quality of services

offered to patients.

2 OUTLINE OF THE E-CARE

PROJECT

Despite considerable advances in medical therapy,

heart failure remains a substantial burden of

mortality and economic cost

(EUROPEAN.COMMISSION, 2012). These trends

underline the growing fiscal and medical imperative

to develop better strategies to improve care delivery

to heart failure patients and reduce rehospitalization

rates.

The healthcare reform in many countries

generates new approaches to care delivery and to

provide high quality. The main reflect is the need for

improving the access of a growing aging population

in order to contain the costs (Zannad et al., 2009).

At the present, the societal and economic

benefits from wider use of telemedicine are far from

being achieved (Weinstein et al.). Then politicians

and healthcare leaders are realizing that telemedicine

is clearly a buttress of the solution. This is tangibly

seen by the soaring number of healthcare systems

that are adopting telemedicine, by the development

of industry investments in telemedicine products and

involvement of government in project delivery.

There common goals is to initiate the citizens on

keeping them healthy, partly by encouraging them to

become more active participants in their own health

management.

Telemedicine provides healthcare services

through use secure transmission of medical data and

Bouguila Z., Moukadem A., Dieterlen A., Ahmed Benyahia A., Hajjam A., Talha S. and Andres E..

Autonomous Cardiac Diagnostic based on Synchronized ECG and PCG Signal.

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

information for the diagnosis, prevention, treatment

and following up patients focusing in benefits of

advanced current technologies and the advance in

signal processing field.

Heart failure (HF) is among the major causes of

hospitalization for elderly citizens (Zannad et al.,

2009). The New York Heart Association (NYHA)

(www.heart.org) functional classification of HF is

widely used classification system relating HF

symptoms. The NYHA classification consists of four

categories that range from no symptoms (class I) to

severe at rest (class IV).

The E-care project’s, through telemedicine, main

objective is to greatly contribute to the enhancement

of the remote patient monitoring, expanding the

possibilities for lifesaving care. In this project, a

smart platform is adopted for home monitoring using

noninvasive sensors to HF patients with NYHA

(class III) severity. This platform is open and

extensible to integrate data sources, to complete the

patient knowledge which will help in diagnostic and

report the risk at an early stage.

Our contribution in this project, with our

partners, is to enhance the signal processing part for

accurate diagnostic. Although ECG and PCG signals

play important roles in heart abnormality detection,

diagnosis based alone on ECG signal or PCG signal

cannot detect most cases of heart symptoms. Hence,

some research has focused on diagnosing heart

defects based on the intercourse between ECG and

PCG signals (Jabloun et al., 2013); (Phanphaisarn et

al., 2011); (Ahlstrom et al., 2008) which can bring

high performance heart diagnostics.

Unlike the time or frequency methods, the time-

frequency analysis has the advantage of being robust

in heart sounds segmentation (Moukadem et al.,

2013). Thanks to this advantage, we want during this

thesis to expand the scope applications to ECG

signals and see also to find common features

between the two signals.

3 HARDWARE AND DATA

ACQUISITION

Chosen sensor are from the market and validated

“Continua” to ensure the compatibility with E-care

platform. A lot of new system will come on the

marquet on the feature, but we focus our attention

only on commercialized product so we work with

aquared signal from system that could be used by

cardiologist.

A laptop will be used as data acquisition for the

proposed analysis. A measurement campaign will be

carried out in the cardiology department of the HUS

of Strasbourg. Heart sound will be captured using

the electronic stethoscope (Littmann Electronic

Stethoscope Model 3200), figure 1. An ECG (éolys),

figure 2, amplifier circuit will be used to capture and

amplify the ECG of patients. Both are wirelessly

connected to the laptop.

3.1 Phonocardiogram Signals

Phonocardiogram (PCG) is the acoustic recording of

mechanical activity of the heart. It facilitates the

measurement of the instantaneous heart rate, beat-to-

beat differences and duration of systolic and

diastolic phases. These measures provide

information about the cardiac function.

Figure 1: 3M™ Littmann® Electronic stethoscope model

3200.

The Littmann electronic stethoscope, figure 1, is

intended for medical diagnostic purposes. It is used

for the detection and amplification of internal sounds

in human body such as from the heart, arteries,

veins, and other internal organs using selective

frequency ranges. It is designed to be used by

anyone who wishes to listen to a sound which is

known, in medical terms, as auscultation.

Using its Bluetooth wireless link, the stethoscope

exchange audio data with an external device in real

time, permitting their visual presentation, recording,

and analysis by applications software.

3.2 Electrocardiogram Signals

The heart produces tiny electrical impulses which

spread through the heart muscle to make the heart

contract.

Figure 2: Electrogradiogram éolys®.

AutonomousCardiacDiagnosticbasedonSynchronizedECGandPCGSignal

The éolys® electrocardiogram (ECG), figure 2,

record the electrical activity of the heart from

electrodes on the body surface. To measure the rate

and regularity of heartbeats, the ECG includes 12

self-adhesive electrodes attached to selected

locations of the skin on the arms, legs and chest.

There is for wirelessly transmitting a 3-/6- or 12-

channel ECG to a monitor, e.g. a PC or a regular

patient monitor.

4 SIGNAL PROCESSING

METHODS

In this experiment, the two physiological signals

(ECG and PCG) will be collected simultaneously but

without electronically controlled synchronization of

the measures.

Advanced methods and techniques of signal

processing and artificial intelligence will be applied

to extract relevant features, after the acquisition,

from the two physiological signals. These signals are

non-stationary by nature. The classical Fourier

transform analyzes the frequency content of signal

without any time information. Therefore, the

importance of time-frequency transforms to detect

the frequency changes of signal over time and to

extract pertinent features form the two physiological

signals.

4.1 S-Transform Challenges

In recent years, joint time and frequency

representation provide a better description of signals

in time-frequency planes. Therefore, the time-

frequency analysis for non-stationary signals is of

great interest and importance in evaluation of signal

characteristics. Mathematical tools of time-

frequency analysis include short-time Fourier

transform (STFT), Wigner-Ville distribution

(WVD), wavelet transform (WT) (Daubechies,

1990) and recently Stockwell Transform (S-

Transform) (Stockwell et al., 1996). S-Transform

leads to multiresolution signal processing, which is

considered as a variable sliding window STFT or as

phase corrected WT.

Stockwell et al., introduced in 1996 the S-

Transform. It combines the potential of the Short

Time Fourier Transform (STFT) and continuous

wavelet transform (CWT) and provides an

alternative approach to process the non-stationary

signals. It employs a moving and scalable localizing

window length. The frequency dependent window

function produces sharper time localization at higher

frequencies and higher frequency resolution at lower

frequencies. Furthermore, the S-Transform has an

advantage, even at the presence of high level of

noise (Stockwell et al., 1996,); (Mansinha et al.,

1997), in that it provides multi-resolution analysis

while it is capable of obtaining reasonably accurate

amplitude and phase spectrum of the analyzed

signals.

The S-Transform of a time series



ht is defined

 



,expexp2

Sf ht iftdt

























(1)

where



exp













is the Gaussian

modulation function and







is the standard

deviation.

where f is the frequency, t and



are both time.

The continuous wavelet transform



,Wd



of a

function





ht is defined as

  

,,Wd htwt ddt











(1)

The S-Transform of the function



ht can also be

defined as a wavelet transform with a specific

mother wavelet multiplied by a phase factor







 

,,exp2SfWd if







(2)

where the mother wavelet



,wtd is defined as

 

,expexp2

ftf

wt f i ft





 





(3)

where the dilation d function is the inverse of f. the

inverse of S-Transform is given by

    

,exp2ht S f d i ftdf

 

 

 











(4)

and, since





,Sf



is complex, can be written as











,,exp,SfAf i f





(5)

where







and







are the amplitude and the

phase of the S-spectrum respectively. The phase

spectrum is an improvement on the wavelet

transform in that the average of all the local spectra

does indeed give the same result as the Fourier

transform.

BIOSTEC2014-DoctoralConsortium

The S-Transform is a useful time-frequency

analysis algorithm. However, it still suffers from

poor energy concentration for the most classes of

signals. An optimization to the existing S-Transform

can enhance the energy concentration in the time-

frequency domain.

In this perspective, as part of my thesis, the first

line of work can be articulated on two theoretical

approaches presented in the following two

paragraphs.

4.1.1 Windows Width Algorithms

Modification of the window width of the S-

Transform enhances the energy concentration in the

time-frequency domain. Djurović et al., (2008)

proposed an algorithm to optimize the window width

in the S-Transform based on the measure of

concentration (Stanković, 2001), which

quantitatively evaluates the energy concentration.

Sejdić et al., (2007) use the Kaiser windows for

improving the energy concentration of the S-

Transform.

4.1.2 Time-frequency Reassignment and

Synchrosqueezing

The Heisenberg uncertainty principle limits the

resolution that can be attained in the time-frequency

plane; different trade-offs can be achieved by the

choice of time-frequency family transform, but none

is ideal. Then the representation can influence the

interpretation given on the time-frequency plane in

order to deduce properties of the signal.

To overcome this difficulty, new techniques,

reassignment (Auger and Flandrin, 1995) and

synchrosqueezing (Daubechies and Maes, 1996), are

recently emerged as a powerful signal processing

tool in non-stationary signal processing. Its basic

objective is to provide a sharper representation of

signals in the time-frequency plane and extract the

individual components of a non-stationary multi-

component signal. These techniques are widely used

in several of new domains, such as audio (Fulop and

Fitz, 2006), physics (Kotte et al., 2006), ecology

(Dugnol et al., 2007), or physiology (Auger et al.,

2013).

4.2 Features Extraction by

Time-frequency Correlation

Some heart problem know as mitral stenosis,

manifested through a heart sound known as the

Opening Snap (OS), is very similar to the third heart

sound (S3). Then, it is very difficult to distinguish

these two sounds without going through proper

training (Erikson, 1997).

In the E-care project we are interested by

detecting the fourth heart sound (S4). The fourth

heart sound is a low-pitched sound and it occurs

shortly before the first heart sound that makes it

detection difficult. For the purpose to study the

fourth heart sound we explore two methods based on

correlation with optimized S-Transform. These

methods can make the detection most effective.

The first method includes a cross-spectral

analysis to study the source localization and phase

synchrony of non-stationary signals (Assous and

Boashash, 2012); (Stockwel, 2007). Since the S-

Transform localizes spectral components in time, the

cross correlation of specific events gives the phase

difference and the amplitude of the cross S-

Transform indicates coincident signals. As the local

phase information can be extracted from the S-

Transform, we can use the cross S-Transform

function to analyze the in-phase and the out-of-phase

components in time-frequency space. This is a very

useful characteristic for cross-spectral and phase

synchrony.

The second method consists on pattern

recognition. The basic idea is to correlate the signal

being analyzed with known template and make

decisions based on the magnitude of the correlation

coefficients, which is between 0 and 1. The process

of correlation is essential to determine the degree of

similarity between the signal being analyzed and the

template. A proposed scheme (Sejdic and Jin, 2007)

known as Selective Regional Correlation (SRC) has

been developed for band limited nonstationary

signals. The preprocessing is carried out by

converting a one-dimensional (1D) time-domain

signal into a two-dimensional (2D) time-frequency

domain representation. The redundant representation

of a (1D) signal in a (2D) time-frequency domain

can provide an additional degree of freedom for

signal analysis to overcome the intertwined time

domain features of the signal and allowing more

importance in the time-frequency domain, resulting

more effective pattern matching.

5 EXPECTED OUTCOME

Using time-frequency reassignment,

synchrosqueezing and correlation function could

makes detection method more effective and accurate

in complex condition with heavy background noise.

In order to perform the knowledge of the heart

activity for automated heart diagnosis and heart

AutonomousCardiacDiagnosticbasedonSynchronizedECGandPCGSignal

disease investigation, we think that using two

physiological signals (ECG and PCG) could be

efficient. Both ECG and PCG signals can thus be

used together for early stage detection of heart

disease into E-care platform.

In this work, an automated system for

preliminary heart defect detection is proposed. This

system is based on the concept of time-frequency

signal analysis.

REFERENCES

Ahlstrom, C., Lanne, T., Ask, P. & Johansson, A. 2008. A

Method For Accurate Localization Of The First Heart

Sound And Possible Applications. Physiol Meas, 29,

417-28.

Assous, S. & Boashash, B. 2012. Evaluation Of The

Modified S-Transform For Time-Frequency

Synchrony Analysis And Source Localisation. Eurasip

J Adv Signal Process, 2012, 49.

Auger, F. & Flandrin, P. 1995. Improving The Readability

Of Time-Frequency And Time-Scale Representations

By The Reassignment Method. Signal Processing,

Ieee Transactions On, 43, 1068-1089.

Auger, F., Flandrin, P., Yu-Ting, L., Mclaughlin, S.,

Meignen, S., Oberlin, T. & Hau-Tieng, W. 2013.

Time-Frequency Reassignment And

Synchrosqueezing: An Overview. Signal Processing

Magazine, Ieee, 30, 32-41.

Daubechies, I. 1990. The Wavelet Transform, Time-

Frequency Localization And Signal Analysis.

Information Theory, Ieee Transactions On, 36, 961-

1005.

Daubechies, I. & Maes, S. 1996. A Nonlinear Squeezing

Of The Continuous Wavelet Transform Based On

Auditory Nerve Models. Wavelets In Medicine And

Biology, Crc Press, 527–546.

Djurović, I., Sejdić, E. & Jiang, J. 2008. Frequency-Based

Window Width Optimization For -Transform. Aeu -

International Journal Of Electronics And

Communications, 62, 245-250.

Dugnol, B., Fernández, C., Galiano, G. & Velasco, J.

2007. Implementation Of A Diffusive Differential

Reassignment Method For Signal Enhancement: An

Application To Wolf Population Counting. Applied

Mathematics And Computation, 193, 374-384.

Erikson, B. 1997. Heart Sounds And Murmurs: A

Practical Guide.

European.Commission 2012. Ehealth Action Plan 2012-

2020– Innovative Healthcare For The 21st Century.

Brussels.

Fulop, S. A. & Fitz, K. 2006. Algorithms For Computing

The Time-Corrected Instantaneous Frequency

(Reassigned) Spectrogram, With Applications. J

Acoust Soc Am, 119, 360-71.

Jabloun, M., Ravier, P., Buttelli, O., Ledee, R., Harba, R.

& Nguyen, L. D. 2013. A Generating Model Of

Realistic Synthetic Heart Sounds For Performance

Assessment Of Phonocardiogram Processing

Algorithms. Biomedical Signal Processing And

Control, 8

, 455-465.

Kotte, O., Niethammer, M. & Jacobs, L. J. 2006. Lamb

Wave Characterization By Differential Reassignment

And Non-Linear Anisotropic Diffusion. Ndt & E

International, 39, 96-105.

Mansinha, L., Stockwell, R. G. & Lowe, R. P. 1997.

Pattern Analysis With Two-Dimensional Spectral

Localisation: Applications Of Two-Dimensional S

Transforms. Physica A: Statistical Mechanics And Its

Applications, 239, 286-295.

Moukadem, A. 2011. Segmentation Et Classification Des

Signaux Non-Stationnaires : Application Au

Traitement Des Sons Cardiaque Et A L'aide Au

Diagnostic. Université De Haute Alsace - Mulhouse.

Moukadem, A., Dieterlen, A., Hueber, N. & Brandt, C.

2013. A Robust Heart Sounds Segmentation Module

Based On S-Transform. Biomedical Signal Processing

And Control, 8, 273-281.

Phanphaisarn, W., Roeksabutr, A., Wardkein, P.,

Koseeyaporn, J. & Yupapin, P. 2011. Heart Detection

And Diagnosis Based On Ecg And Epcg

Relationships. Med Devices (Auckl), 4, 133-44.

Ping, Z. & Zhigang, W. A Computer Location Algorithm

For Ecg, Pcg And Cap. Engineering In Medicine And

Biology Society, 1998. Proceedings Of The 20th

Annual International Conference Of The Ieee, 29 Oct-

1 Nov 1998 1998. 220-222 Vol.1.

Sejdic, E., Djurovic, I. & Jin, J. S-Transform With

Frequency Dependent Kaiser Window. Acoustics,

Speech And Signal Processing, 2007. Icassp 2007.

Ieee International Conference On, 15-20 April 2007

2007. Iii-1165-Iii-1168.

Sejdic, E. & Jin, J. 2007. Selective Regional Correlation

For Pattern Recognition. Systems, Man And

Cybernetics, Part A: Systems And Humans, Ieee

Transactions On, 37, 82-93.

Stanković, L. 2001. A Measure Of Some Time–Frequency

Distributions Concentration. Signal Processing, 81,

621-631.

Stockwel, R. 2007. Why Use The S-Transform. Ed. By A

Wong (Fields Institute Communications, Providence),

Pseudo-Differential Operators: Pdes And Time-

Frequency Analysis, 279-309.

Stockwell, R. G., Mansinha, L. & Lowe, R. P. 1996.

Localization Of The Complex Spectrum: The S

Transform. Signal Processing, Ieee Transactions On,

44, 998-1001.

Weinstein, R. S., Lopez, A. M., Joseph, B. A., Erps, K. A.,

Holcomb, M., Barker, G. P. & Krupinski, E. A.

Telemedicine, Telehealth And Mhealth Applications

That Work: Opportunities And Barriers.

The American

Journal Of Medicine.

Zannad, F., Agrinier, N. & Alla, F. 2009. Heart Failure

Burden And Therapy. Europace, 11 Suppl 5, V1-9.

BIOSTEC2014-DoctoralConsortium