Models for ECG :

A Dynamical Model for Generating

Synthetic Electrocardiogram Signals

Patrick E. McSharry, Gari D. Clifford, Lionel Tarassenko, and Leonard A. Smith

References SLIDE 2
❑Patrick E. McSharry, Gari D. Clifford, Lionel Tarassenko, and Leonard A.
Smith (2003), Dynamical Model for Generating Synthetic Electrocardiogram
Signals, IEEE Transactions on Biomedical Engineering, Vol. 50, No.3, pp:
289-295.
❑Gari D. Clifford, Francisco Azuaje, Patrick E. McSharry (2006), “Advanced
Methods and Tools for ECG Data Analysis”, Artech House, Norwood.
❑Saeid R. Seydnejad and Richard I. Kitney (2001), Modeling of Mayer Waves
Generation Mechanisms: Determining the Origin of the Low- and Very Low
Frequency Components of BPV and HRV, IEEE Engineering in Medicine and
Biology, pp: 92-100.
Introduction (1) SLIDE 3
❑The availability of open-source computational models and simulators can
greatly facilitate the advancement of cardiovascular research.
❑There is an advantages between the development of these models and the
exploration of biomedical databases obtained from clinical studies :
1. Researchers can construct and evaluate their models/database using the
biomedical signals.
2. The ability to simulate realistic signals using these models/database can
be used to assess novel biomedical signal processing techniques.
3. These models/database can be used to formulate new experimental
hypotheses.
Introduction (2) SLIDE 4
❑The ECG signal describes the electrical activity
in the heart and each heartbeat traces the
familiar morphology labeled by the P, Q, R, S,
and T peaks.
❑The R peak is typically associated with the
largest deflection away from the baseline, this
peak is generally taken as a marker for each
heartbeat as it is the easiest to locate.
❑RR tachogram ➔ RR intervals forms the basis
of the RR interval time series
The Cardiovascular System (1) SLIDE 5

❑P-wave: A small low-voltage deflection away from the baseline caused by the depolarization of the atria
prior to atrial contraction as the activation (depolarization) wave-front propagates from the SA node
through the atria.
❑PQ-interval: The time between the beginning of atrial depolarization and the beginning of ventricular
depolarization.
❑QRS-complex: caused by currents generated when the ventricles depolarize prior to their contraction.
The Cardiovascular System (1) SLIDE 6
❑QT-interval: The time between the
onset of ventricular depolarization
and the end of ventricular
repolarization.
❑ST-interval: The time between the
end of S-wave and the beginning of
T-wave. Significantly elevated or
depressed amplitudes away from the
baseline are often associated with
cardiac illness.
❑T-wave: Ventricular repolarization,
whereby the cardiac muscle is
prepared for the next cycle of the
ECG.
The Cardiovascular System (1)
❑ The autonomic nervous ❑ Effects of sympathetic and parasympathetic activation:
system plays a primarily
role to adjust
parameters of
cardiovascular.
❑ The efferent autonomic
signals are transmitted
to the various organs of
the body through two
major subdivisions :
sympathetic nervous
system and the
parasympathetic
nervous system.
 Respiratory Sinus Arrythmia SLIDE 8
❑Respiration, a mainly parasympathetically
mediated process, is the most obvious
observable phenomenon in the RR
tachogram.
❑The RR interval will oscillate periodically,
shortening with inspiration (and
lengthening with expiration) ➔ This
phenomenon known as Respiratory Sinus
Arrhythmia (RSA).

❑During inspiration, the pressure within the thorax decreases and

venous return increases, which stretches the right atrium resulting
in a reflex that increases the local heart rate (i.e., shortens the RR
intervals).
❑During expiration, the reverse of this process results in a slowing of
the local heart rate.
❑The phase between the respiratory RR interval oscillations and
respiratory-related changes in ECG morphology is not static
 RR SLIDE 9
Tachogram

❑In general, the spectral power in the RR

tachogram is broken down into four bands:
1. Ultra low frequency (ULF): 0.0001 Hz ≥ ULF <
0.003 Hz;
2. Very low frequency (VLF): 0.003 Hz ≥ VLF <
0.04 Hz;
3. Low frequency (LF): 0.04 Hz ≥ LF < 0.15 Hz;
4. High frequency (HF): 0.15 Hz ≥ HF < 0.4 Hz.
Why modelling of ECG signal? SLIDE 10
PhysioNet
❑ECG signal data in PhysioNet sometimes includes no P waves.
❑Sometimes no variations in timing or morphology ➔ short recording
❑The signal is discontinuities.

Synthetic ECG
❑Based upon time-varying differential equations. Possible to
simulate ECG
❑Continuous with convincing beat-to-beat variations in morphology and
signals that
interbeat timing.
show
❑Generate extremely realistic ECG signals with complete flexibility over signs of various
the choice of parameters that govern the structure of these ECG signals pathological
in both the temporal and spectral domains. conditions.
Algorithm for Synthetic ECG SLIDE 11
The algorithm consists of two parts:
1. The generation of an internal time series with internal sampling
frequency fint to incorporate a specific mean heart rate, standard
deviation, and spectral characteristics corresponding to a real RR
tachogram.
2. The second stage produces the average morphology of the ECG by
specifying the locations and heights (amplitudes) of the peaks that occur
during each heartbeat.
Software Block SLIDE 12
Diagram :
Spectral Characteristics (1) SLIDE 13
❑Spectral characteristics of the RR tachogram,
including both RSA and Mayer waves, are
replicated by describing a bimodal spectrum
composed of the sum of two Gaussian
functions :

c1, c2 : standard deviation

σ12, σ22 : power in LF and HF bands
Mayer Waves SLIDE 14
❑The origin of LF and VLF components ➔ Mayer waves.
❑Three oscillatory systems that are responsible for generation of the Mayer
waves :
1. The baroreceptor or chemoreceptor feedback system.
2. Centrogenic rhythm in the brain stem with interconnection to respiratory
oscillator.
3. The autorhythmicity of the vascular smooth muscle.
RR Tachogram Time Series SLIDE 15
❑A time series T(t) with power spectrum S(f) is generated by taking the
inverse Fourier transform with amplitudes
❑By multiplying this time series by an appropriate scaling constant and adding
an offset value, the resulting time series can be given any required mean and
standard deviation.
❑Suppose that T(t) represents the time series generated by RR-process with
power spectrum S(f). The angular velocity of motion around the limit cycle is
then given by :
3D State Space (1) SLIDE 16
❑During each heartbeat, the ECG traces a
quasi-periodic waveform where the
morphology of each cycle is labeled by its
peaks and troughs, P, Q, R, S, and T.
❑This quasi-periodicity can be reproduced by
constructing a dynamical model containing
an attracting limit cycle.
❑Each heartbeat corresponds to one
revolution around this limit cycle which lies in
the (x, y)-plane.
❑A series of five angles (θP, θQ, θR, θS, θT)
describes extrema of the peaks P, Q, R, S, T.
3D State Space (2) SLIDE 17
❑The dynamical equations of motion are given by three ordinary differential eq.:

4th order Runge-Kutta

Method to integrate the eq.

ω : angular velocity of the trajectory

ai : the magnitude of the peaks
bi : width (time duration) of each peak.
❑Baseline wander :
Morphological Parameters (2) SLIDE 18

❑A nonlinear relationship between the morphology modulation factor and

the heart rate decreases the temporal contraction of the overall PQRST.
❑The changes in angular frequency, around the limit cycle, resulting from
the period changes in each RR interval, do not lead to temporal changes,
but to amplitude changes.
Morphological Parameters (1) SLIDE 19
❑The size of the mean heart rate (hmean) affects the shape of the ECG
morphology.
❑The QRS width decreases with increasing heart rate ➔ These changes are
replicated by modifying the width of the exponentials and also the
positions of the angles ➔ This is achieved by using a heart rate dependent
factor
❑Suppose that T wave is asymmetry ➔ replicated by adding an extra
Gaussian to the T wave section (T- and T+ are placed before and after the
peak of T wave in the original model)
Morphological Parameters (2) SLIDE 20
Temporal and
Spectral Parameters
of the ECG Model
(for Normal ECG) :
Morphological Parameters (3) SLIDE 21
❑The realistic amplitude variation
which respect to the electrode
position in real recordings, is
dominated by the high-frequency
component in Gaussian
distribution.
❑The part of the model could be
made more realistic by coupling
the baseline wander to a phase-
lagged signal derived from
highpass filtering. (fc = 0.15 Hz)
the RR interval time series.
Morphological Parameters (4) SLIDE 22
❑Estimation of the LF/HF ratio were calculated for
a range of sampling frequencies (Lomb • LF/HF ratio estimates computed from
periodogram): synthetic ECG signals for a range of
sampling frequencies using an input
LF/HF ratio of 0.5 (horizontal line).
• The distribution of estimates is shown
by the mean (dot) and plus/minus one
standard deviation error bars.
• It may be concluded that the
underestimation of the LF/HF ratio is
due to the sampling frequency being
too small.
Results SLIDE 23
• Observational uncertainty is
incorporated by adding
normally distributed
measurement errors with mean
zero and standard deviation
0.025 mV, yielding a similar
signal to a segment of real ECG
from a normal human
Results SLIDE 24
A comparison between the
continuous process with power
spectrum S(f) and RR-process
obtained from the R-peak
detection ➔ illustrates the
measurement errors that arise
when computing HRV statistics
from RR-intervals.
Results SLIDE 25
Analysis of RR-intervals from R-peak
detection of the ECG signal generated by
the dynamical model (1) with mean heart
rate 60 bpm and standard deviation 5
bpm.
(a) RR-intervals.
(b) Instantaneous heart rate.
(c) Power spectrum of the RR-intervals.
Note the two active frequencies belonging
to RSA (0.25 Hz) and Mayer waves (0.1
Hz).
Results SLIDE 26
Results SLIDE 27
ASSIGNMENT
SLIDE 28
ECG Signal with
premature ventricular
contraction :
That’s all. Thank you very much! ☺
Any Questions?
3D State Space (3) SLIDE 30
❑The output synthetic ECG signal, s(t), is the vertical component of the three-
dimensional dynamical system ➔ s(t) = z(t).
❑Having calculated the internal RR tachogram expressed by the time series T(t) with
power spectrum S( f ) given by Gaussian Eq., this can then be used to drive the
dynamical model 3D state space so that the resulting RR intervals will have the same
power spectrum as that given by S(f).
❑The time interval T(tn) is used to calculate an angular frequency
❑Angular frequency is used to specify the dynamics until the angle θ reaches θR again,
whereby a complete revolution (one heartbeat) has taken place.
❑For the next revolution, the time is updated,
❑The next angular frequency is used to drive the trajectory around the
limit cycle