elevation MI (NSTEMI) is also possible. The ST segment Abstract

ECG Simulation for Myocardial Infarction
Diagnosis in High Fidelity Mannequins
K Kanakapriya, Alekhya Mandali, M.Manivannan*
Abstract—Cardiovascular diseases have reached epidemic
proportions in developing countries, and acute Myocardial
Infraction is the most common manifestation of this disease. A
high fidelity mannequin is being built towards training medical
personnel in diagnosing MI.
This paper describes the
simulation of a 12 lead ECG, the primary tool for recognizing
type and stage of MI, in the mannequin and its incorporation
into the hemodynamic model, ensuring realistic variations in
beat-to-beat timing. The module also verifies correct electrode
placement for ECG.
Keywords-High-Fidelity Mannequin, ECG Simulation,
Cardiovascular Simulation, 12 lead ECG, MI Diagnosis
I. INTRODUCTION
Myocardial Infarction (MI) occurs when the blood supply to
a specific region(s) of the heart is interrupted, damaging or
destroying the heart muscle. Treating MI within hours of
onset of symptoms could minimize cell necrosis in the
affected areas, preserving most of the muscle function.
Acute Chest Pain the primary symptom of MI, may also be
symptomatic
of
cardiac,
vascular,
pulmonary,
gastrointestinal, musculoskeletal, infectious or even
psychological disease. MI can be diagnosed by reviewing
the patient’s medical history, physical condition, genetic
predisposition and Electrocardiogram (ECG) recording.
Cardiac markers like the CK/CKMB enzyme, manifest only
2 – 3 hours after the actual event and these changes are
sometimes non-specific, they occur in damage of skeletal
muscle also. ECG is the quickest physiological record to
record changes in the myocardium because of MI.
ECG is a record of potential changes at the skin surface
because of polarization and depolarization of the heart
muscle, not the actual tiny pacemaker conduction system
[3], since the pacemaker potential is too low to be felt at the
skin. Consequently an abnormal ECG is indicative of
damage to the heart muscle. ECG records the atrial
depolarization – P Wave, the AV node delay to allow atrial
contraction to complete – PR interval, the ventricular
depolarization – QRS complex, the isoelectric rapid ejection
phase – ST interval and the ventricle repolarization – T
Wave. If a region of the myocardium is damaged because of
MI, ST segment is not isoelectric anymore, injury currents
result in ST segment elevation, but non ST segment
M.Manivannan is with Biomedical Engineering Group, Department of
Applied Mechanics, Indian Institute of Technology Madras, Chennai600036, India (corresponding author phone: 091-44-22574064; e-mail:
[email protected]).
K. Kanakapriya is with Biomedical Engineering Group, Department of
Applied Mechanics, Indian Institute of Technology Madras, Chennai600036, India (e-mail: [email protected]).
elevation MI (NSTEMI) is also possible. The ST segment
elevation occurs at the start of heart attack and lasts a few
hours. Q Wave changes and T wave changes mark the onset
of necrosis. Pathological Q Waves are remnants of a MI and
are evidence of dead muscles inside the myocardium.
It is essential to diagnose MI and start appropriate therapy,
whether reperfusion or clot dissolving anti-thrombolytics, to
ensure further physiological damage does not occur in other
vital organs and MI does not result in other cardiac
complications. Reperfusion of blocked cardiac arteries
minimize damage to the myocardium allowing the patient to
return to a normal life quickly, even otherwise, diagnosing
MI prepares the medical staff to watch out for change in
heart rhythm or blood pressure which could signal further
damage to the myocardium.
Cardiac Heart Disease occurrence in South Asia has become
a health epidemic [1] and the age at which Acute Myocardial
Infarction, the primary manifestation of CHD, occurs, is
falling (56 years) [2]. It has become imperative to ensure
medical personnel are able to quickly diagnose MI to
embark on further timely treatment. A high-fidelity
mannequin which replicates human body anatomy and
physiology relevant to MI, respond to relevant treatment or
intervention and is able to supply objective data regarding
resident actions through debriefing software would make it
easier to acquire, retain and test such specialized skills.
ECG Module is not the focus of most high-fidelity
mannequins, ECG electrode placement and signal
interpretation are usually taught separately; the ECG signal
is generated from a database of ECG recordings or by
modifying standard waveforms, ignoring RR interval
variability. This paper attempts to combine the ECG
placement hardware and a realistic ECG Synthesis software.
II. METHODOLOGY
The typical single lead ECG only looks at the frontal plane
of the heart, MI Diagnosis requires a 12-lead ECG, 4 limb
electrodes and 6 chest electrodes combining to give a three
dimensional view of the heart. This ensures even localized
changes because of MI or ischemia are not missed. ECG
Simulation for a high-fidelity mannequin has three major
components, placing the leads correctly to generate ECGs,
synthesizing the 12 lead patient specific ECG and
incorporating the ECG Module into the cardiovascular
hemodynamic simulation.
A. ECG SIMULATION HARDWARE
In clinical practice, the 12 lead ECG requires the 10
numbered electrodes to be placed at specific locations, 2 on
the arms, 2 on the legs and 6 on the chest. Incremental
adjustments in position may be carried out with feedback
from the ECG monitor to get a good quality signal. Interchanging the lead positions will not throw up any errors in
the ECG monitor, but would result in definite misdiagnosis
from the ECG Recording. Hence it follows, two inputs are
necessary to determine correct lead placement, the electrode
number and the electrode position on the mannequin.
PROCESSOR
PORT
FSR BASED
TRIGGER
PROCESSOR
PORT
FSR BASED
TRIGGER
MEMBRANE
POT. CKT
RESISTANCE
TO
FREQUENCY
PROCESSOR
PORT
DISPLAY ERROR
Potentiometers. The resistance activated when the electrode
is attached is mapped to a specific frequency which maps to
a specific chest location. As detailed in Fig. 1, the ECG is
displayed only when the electrode and mannequin ECG
specific electrode locations match.
B. ECG MORPHOLOGY SIMULATION
Two disparate approaches have been implemented for
including a 12-lead ECG in the MI diagnosis mannequin,
one is using canned ECG from the Physikalisch-Technische
Bundesanstalt (PTB) Diagnostic ECG database [4] and the
other is statistical model implementation based on an
empirical description of ECG [5].
The PTB Database is a compilation of 549 digitized ECG
records from 290 subjects, some healthy and some with
different heart conditions. The ECG data, corresponding to
the disease condition being studied, for a single heart beat in
various heart rate ranges has been stored. This data is scaled
to match the instantaneous heart rate of the mannequin,
obtained from the hemodynamic model, and displayed. The
amplitude of the ECG morphology is not altered.
It is possible to specify the ECG morphology for a specific
type or stage of heart attack from medical literature.
ECGSYN [6] is a dynamical model for generating ECGs
with arbitrary morphologies allowing control over the
structure of the ECG in both temporal and spectral domains.
In this implementation the hemodynamic model provides the
expected beat to beat variability from interventions and/or
MI progression, so only the ECG Morphology synthesis part
of the model has been adapted.
In ECGSYN the quasi-periodicity of the ECG is reproduced
by a trajectory along an attractor circle of unit radius in the
X-Y plane. The specific P, Q, R, S, T morphology of the
ECG is generated by forcing the trajectory, towards or away
from the attractor circle along the Z-axis, at specific angular
positions. The peaks and troughs along the Z-axis are
modeled as Gaussians.
x = α x − ω y
y = α y + ω x
No
IDENTIFY
ELECTRODE
NUMBER
Match
Yes
DISPLAY ECG
IDENTIFY
ECG
LOCATION
COMPUTER
Fig. 1: The electrodes and the placement sites activate different ports
in the processor, the computer verifies the match and allows ECG to
be displayed.
The mannequin electrodes are provided with Force Sensitive
Resistors (FSR) which are activated when pressed upon the
mannequin and map to a specific port addresses in the
microprocessor, thus identifying the electrode number. The
limb electrode positions are also provided with FSRs, which
are activated when the electrode is pressed upon them and
they also map to specific addresses on the processor. The
chest electrode positions are provided with Membrane
z = −
∑
i∈( P ,Q , R , S ,T )
α i Δθi exp(−Δθi2 / 2bi2 ) − ( z − zo )
Where α = 1 - √x2+y2, ∆θi = (θ-θi) mod 2π, θ = atan2(y/x)
and ω is the angular velocity of the trajectory around the
circle. The Gaussian coefficients αi determines the
magnitude of the peaks while the coefficients βi determine
the time duration of the peaks. Baseline wander is added by
coupling the baseline value zo to respiratory frequency fr
using zo(t) – Asin(2πfr). The output ECG curve is the
vertical component, z(t) of the three dimensional trajectory.
In this implementation, the angular positions along the
trajectory and the amplitude and height of the Gaussians
corresponding to the P, Q, R, S, T waves are determined
empirically from ECGs corresponding to the specific MI
type being simulated. The PTB database ECGs have been
used to create these specification. The angular velocity ω is
determined from the beat-to-beat time of the hemodynamic
model. 12 different ECG waveforms corresponding to 12
different leads are synthesized for each heartbeat. Figure 2
shows a sample ECG from leads I and II following Inferior
MI.
Voltage mV
C. ECG INTEGRATION IN CVSIM
CVSIM consists of 6 compartments, left ventricles, right
ventricles, pulmonary arteries and veins, systemic arteries
and veins giving rise to 6 coupled ordinary differential
equations. The ventricles are modeled by a time varying
capacitor; the other compartments are each modeled by a
Time s
Lead I
Lead
Lead III
Lead avF
Fig. 2: ECG Synthesized for Leads I, II, III and aVF. Lead I displays a normal ECG, Lead II shows T-Wave inversion, Lead III and Lead
aVF show T – Wave inversion and an abnormal Q-Wave.
Treatment options in any Intensive Care Unit are based upon
monitored hemodynamic parameters like the ECG, System
Arterial Pressure, Central Venous Pressure, Radial Artery
Pressure, Respiration rate Heart Rate and Blood Pressure.
Lumped parameter approximations of the distributed
cardiovascular systems generate hemodynamic waveforms
that are reasonable approximations of those obtained from
the human cardiovascular system. The cardiovascular
simulator CVSIM [7] is one such dynamic simulator of the
lumped parameter model of human cardiovascular
hemodynamics
linear capacitor and a linear resistor. The ODEs are solved
by numerical integration using 5th order Runge Kutta method
with adaptive step size.
Cardiovascular regulation, occurring at both extrinsic global
level and intrinsic local level, aims at maintaining
homeostasis with multiple feedback (Arterial Blood
Pressure, Right Atrial pressure) and controls. When short
term responses are considered (within seconds) the Arterial
Baroreflex system and the Cardiopulmonary Baroreflex
system, both mediated by the Autonomous Nervous System
play the principal role in extrinsic control of hemodynamic
ABR
Csp,esl,r
Contractility
ANS
ABR
Heart rate
Cesl,r
SA Node
HEART
Psp
Pa(t)
Pra(t
CIRCULATION
ECG Synthesis
CBR
Venous
CBR
ABR
Venous
CBR Static
Saturation
V0
Arterial Resistance
Rspa
Pspra
ABR Static
Saturation
Ra
ABR
Arterial Resistance
Vsp0
Fig. 3: The diagram shows how the Arterial and Cardiopulmonary Baroreflex modulate the hemodynamic parameters and launch the ECG
synthesizer for each heartbeat. The current pressure value (Aortic or Right Atrial) is compared with the corresponding set point and the difference
is scaled to limit error signal and this signal is sent to the ANS. The ANS determines the contribution of the signal to each of the four
hemodynamic parameters, Heart Rate, End-Systolic Capacitance of the left and right ventricles, Arterial Tone and Venous Tone.
parameters through the modulation of Total Peripheral
Resistance, Venous tone, Ventricular Contracility and Heart
Rate. Fig. 3 shows a block diagram explaining the flow of
feedback and control.
In the Arterial BaroReflex (ABR) Mechanism, stretch
receptors in the aortic arch and carotid sinus respond to
changes in Arterial Blood Pressure (ABP) causing electric
signals to travel along afferent fibres to the centre of
automatic activity in the medulla oblongata. This leads to
reciprocal effects on two efferent limbs of the Automatic
Nervous System, increased afferent nerve traffic causes
decrease in efferent sympathetic outflow and increase in
parasympathetic outflow.
Sympathetic outflow causes
increase in heart rate, increase in cardiac contractility,
increase in arterial resistance and decrease in zero pressure
filling volume. Parasympathetic outflow causes a decrease
in heart rate. Cardiopulmonary reflex mechanism also
behaves similarly with stretch receptors in the atria-caval
junctions, atrial and ventricular walls. In the CVSIM
implementation of ABR, the moving average of aortic
pressure generated by the hemodynamic model is compared
with an arterial set point pressure. The difference is scaled
to limit the error signal to +/- 28 mm Hg. This signal is
convolved with Impulse response function for sympathetic
and parasympathetic actions and scaled by corresponding
static gain values to determine the contribution of the ABR
to the effector variable X[n]. Changes in arterial resistance
and venous tone occur smoothly by interpolating the discrete
values of X[n], changes to contractility is affected at
beginning of the next beat, the time interval after which the
next beat occurs is modulated by the effector variable as
described in the next section. Cardiopulmonary Baroreflex
(CBR) implementation is very similarly to the ABR, the
sensed variable is the right atrial pressure, after subtracting
from a cardiopulmonary set point, the scaled error signal is
convolved with impulse functions and multiplied by static
gain values to determine contribution to the effector
variable. However, CBR has no influence on cardiac
contractility or heart rate.
In the cardiovascular system the SinoAtrial (SA) node serves
as the pacemaker. This node, richly enervated with
sympathetic and parasympathetic nerve fibres, modulates the
heart rate in accordance with the ANS mechanisms. The SA
node is modeled as a function whose value at any time
depends on the cumulative contributions of automaticity and
autonomic activity since the last cardiac firing. Once the
function reaches a threshold value and crosses a minimum
refractory time (one fifth of the previous cardiac cycle time),
it fires again, generating a heartbeat and resetting the
function value to zero once again. The ECG signal is
generated simultaneously, ensuring the hemodynamic model
timing matches the electrical timing.
III.
DISCUSSION
CVSIM has been modified to run the backend hemodynamic
simulator for the MI mannequin. Instead of launching the
simulator after setting a table of parameters like heart rate,
contractility, peripheral resistance, the simulator is launched
by selecting a specific patient. Each patient is mapped to a
specific MI Scenario, currently only Inferior Myocardial
Infarction, which is caused by an occlusion in the right
coronary artery and occurs at the base of the left ventricle.
Inferior MI is diagnosed from changes in leads II, III and
aVF as described in Fig. 2. The morphology of inferior
infraction is well documented and has been used to generate
the morphology shown. Using canned ECGs in the simulator
would be more realistic, allowing minor unaccounted
changes in the morphology; Using the synthesized ECG
allows the resident to focus just on the changes specific to
the MI type being studied. This simulator has provisions for
both types.
The cardiac pacemaker, SA Node
implementation in CVSIM, allows the ANS to modulate the
default heart rate by changes in the homeostasis because of
MI or drug interventions to treat MI. This also generates a
realistic heart rate rhythm in the ECG. The simulator tracks
the speed and suitability of the interventions to determine the
outcome.
The current implementation does not model any change in
atrial or ventricular contraction timing. The systolic time is
always one third the square root of the cardiac cycle time.
The isovolumetric relaxation time is always half of the
systolic time. There can be only inter-beat match between
the cardiac cycle and the ECG cycle in this implementation.
Intra-beat match would require coupling the preset ECG
morphology time to the systolic and diastolic time
predefined in the cardiovascular model. It is therefore not
possible to dynamically simulate ventricular fibrillation, a
possible consequence of acute MI with this model.
There is no dynamic coupling between the mechanical
model of the heart, the cardiovascular cycle and the
electrical model of the heart, of which ECG is a snapshot;
They only occur simultaneously. This implementation only
looks at the short term control of the Cardiovascular system,
hormonally mediated extrinsic control which regulates ABP
over longer periods (hours to days) is not considered. Only
high pressure side of ABP control has been considered,
Arterial Chemoreflex System [8], which becomes active
when Arterial Blood Pressure falls below 80 mm Hg is not
modeled, though such conditions could arise after Inferior
MI.
IV. SUMMARY
Here a 12 Lead ECG generation module has been
implemented for a high-fidelity mannequin to diagnose and
treat Acute Myocardial Infarction. CVSIM, a lumped
parameter models the cardiovascular system has been
integrated with the ECG generation module, but while the
ECG matches the cardiovascular cycle in the time scale,
displaying realistic beat to beat time variations, it is a
statistical model based on empirical description of the ECG
for each particular condition.
The mechanical
cardiovascular model is not coupled to the electrical model
of the heart. However, it provides enough data to diagnose
MI and make appropriate treatment choices. The module
also has the additional novelty of verifying correct electrode
placement.
A 3-dimensional model of the heart, integrating the
mechanical and electrical models and generating a forward-
model of ECG would be the ideal simulator for MI. The
electrical model could be a single dipole model or an
equivalent double model [9], then, the ECG Morphology
would automatically match the MI-Type instead of being
synthesized independent of the hemodynamic system. The
ANS needs to be expanded to include arterial chemoreflex
controls and the hemodynamic model should be expanded to
include coronary circulation.
ACKNOWLEDGMENT
We greatly appreciate the funding from Department of
Biotechnology, Government of India for undertaking this
project. We also would like to thank colleagues on the other
First Aid mannequin projects, Mr Sathish Kumar G, Mr
Prasanna Gandhiraj and Mr Karthik for their valuable input.
REFERENCES
1. Gupta R. (2007). "Escalating Coronary Heart Disease and z
Factors in South Asians" . Indian Heart Journal: 214–17.
2. P Joshi, S Islam, P Pais, S Reddy, et al., “Risk Factors for Early
Myocardial Infarction in South Asians Compared With
Individuals in Other Countries”, JAMA, Pp, 286 – 294, Vol.
297 No 3, Jan 17 2007
3. J.R.Levick, ”An introduction to Cardiovascular Physiology”, Pp
58 – 68, Arnold Viva Edition 2005
4. Bousseljot, R.; Kreiseler, D.; Schnabel, A. Nutzung der EKGSignaldatenbank CARDIODAT der PTB über das Internet.
Biomedizinische Technik, Band 40, Ergänzungsband 1 (1995) S
317
5. McSharry PE, Clifford GD, Tarassenko L, Smith L. A dynamical
model for generating synthetic electrocardiogram signals. IEEE
Transactions on Biomedical Engineering 50(3): 289-294; March
2003
6. P. E. McSharry and G. D. Clifford, ECGSYN - A realistic ECG
waveform generator. [Online]. Available:
http://www.physionet.org/physiotools/ecgsyn/
7. T Heldt, EB Shim, RD Kamm, and RG Mark. Computational
modeling of cardiovascular response to orthostatic stress. J Appl
Physiol 92(3):1239-1254 (2002 March 1)
8. T Heldt, R Mukkamala, GB Moody, and RG Mark. CVSIM a
opensource cardiovascular simulator for teaching and research.
The Open Pacing, Electrophysiology and Therapy journal, 2010,
3 :45-54
9. A Van Oosterom, T F Oostendorp, ECGSIM: an interactive tool
for studying the genesis of QRST waveforms, Heart, HT14662,
2003