Semi-Blind Source Separation for fetal ECG extraction

Introduction 
 
1 
 
1 INTRODUCTION 
The electrocardiogram (ECG) signal is one of the most commonly employed 
physiological signals. ECG signal obtained by measuring the differences of potential 
generated by electrical activity of the heart through electrodes placed generally on the 
skin of the patient. The ECG can provide valuable clinical information. The monitoring 
and analysis of the ECG is a useful technique to support diagnosis of many cardiac 
diseases. There are two main significant features of the ECG useful in clinical 
diagnoses: its morphology and its timing (R-R interval is the reciprocal of the heart 
rate). 
As in adults, the well-being of a fetus can be assessed from the fetal ECG (fECG) signal 
during pregnancy. The fECG signal can be obtained from electrical measurements on 
the maternal abdomen. However, the abdominal ECG signal consists of a combination 
of the maternal ECG (mECG) signal, the fECG signal, and interference signals. As the 
amplitude of the mECG signal is typically much larger than the fECG signal and the 
interference signals can be comparable to fECG amplitude, fetal electrocardiography 
has been disregarded in clinical application for years. Signal processing can offer an 
effective solution to provide reliable fECG signals, also for clinical application. 
Nowadays due to the difficulty for the extraction of the fECG, a technique widely 
accepted to assessing the conditions of well-being of the fetus is the registration of the 
Fetal Heart Rate (FHR). 
In particular, the Doppler Ultrasound Cardiotocography (US-CTG) is the most 
widespread clinic prenatal methodic to assess fetal well-being. Nowadays, the fECG is 
considered as a possible alternative or complementary technique to the US-CTG. 
In order to obtain more and more detailed information on the well-being of the fetus is 
important to make more frequent and long-lasting FHR recordings; but the Doppler 
Ultrasound Cardiotography is an active technique, i.e. involves an administration of 
energy and it may create some problems for long-term monitoring. 
It is also worth mentioning that the US-CTG only provide fetal heart rate information 
and this it is not enough to provide certain diagnosis of some ominous fetal states. 
Indeed, fECG can provide more information than the mere heart rate by analyzing its 
morphology.
Introduction 
 
2 
 
In this research study, I tried to find new solutions for processing the fECG signals 
taken by maternal abdomen. The fECG is important for getting information for 
diagnosis and for basic research, in fact the study of the morphology of the fECG may 
add more information to the classical techniques, e.g. could provide clear evidence of 
the condition of the fetus (e.g. hypoxia, acidemia) allowing physician to provide timely 
appropriate therapies. It is also important to remember that this is a passive technique; 
therefore it allows obtaining long-lasting and frequent patient monitoring. 
This work is based on the application of innovative techniques of processing of signals, 
such as the Independent Component Analysis (ICA), on a set of recordings of electrical 
potential obtained with various electrodes placed on maternal abdomen and a priori 
knowledge of the noises. 
ICA is able to separate the fetal signal from the maternal and other noises, but in order 
to work well it needs a very high number of ECG abdominal leads, which is unpractical 
for clinical applications. 
For this reason, I used a Semi-Blind approach based on a combination of noise 
suppression using a priori knowledge and a successive application of ICA methods. In 
this way, I was able to correctly extract the fECG signal by using only four abdominal 
leads. 
The chapters of this dissertation are organized as follows. The 2
nd
 chapter introduces the 
problem of fetal monitoring, in particular the problem of fetal electrocardiography. The 
3
rd
 chapter presents a review of various methods utilised for extracting the fECG. In the 
4
th
 chapter I explained the Blind Source Separation (BSS) techniques and in particular I 
dealt with second and higher order statistical computation, i.e. Principal Component 
Analysis (PCA) and Independent Component Analysis (ICA). Experimental results, 
with the application of two ICA algorithms on a set of real signals, are demonstrated in 
the 5
th
 chapter. Finally the conclusions and further work are discussed in the last 
chapter.
Fetal Monitoring 
 
3 
 
2 FETAL MONITORING 
2.1 A Brief History 
Fetal Heart Sounds (FHS) were reportedly first detected by Marsac in the 1600‟s. The 
idea that Fetal Heart Rate (FHR) could be used to determine fetal well being was first 
proposed by Killian in the 1600‟s. This idea went unnoticed until 1818 when Mayor and 
Kergaradec described the method of auscultating fetal heart sounds by placing the ear 
next to the maternal abdomen. By 1833, Evory Kennedy, an English physician, 
published guidelines for fetal distress and recommended auscultation of the fetal heart 
rate. In 1893, Von Winkel established criteria for fetal distress that remained unchanged 
until the advent of electronic fetal monitoring (tachycardia – FHR>160, bradycardia – 
FHR<100, irregular heart rate, passage of meconium, and gross alteration of fetal 
movement) [1]. 
 
Figure 2.1 History of Fetal Electrocardiography [1] 
The fetal stethoscope, or fetoscope, was first described by David Hillis in 1917 at the 
Chicago Lying-In Hospital. It remained so until well into the 1970s, and is used in some 
form even today. 
In 1906, Cremer first measured the fECG by using abdominal electrodes. Since Cremer, 
FHR monitoring has been used clinically for assessing fetal health and status. Fetal 
cardiograms can predict fetal distress allowing doctors to prevent irreversible harm to 
the fetus [2]. Over the next 50 years, varieties of improvements to fECG were made in
Fetal Monitoring 
 
4 
 
the way of amplification and abdominal electrode placement, mostly in an attempt to 
improve resolution of the fetal QRS complex and calculate the FHR. 
2.2 Monitoring Techniques 
There are two situations for which fetal monitoring provides important information 
about the condition of the fetus: 
 reactive fetal monitor tracing identifying a fetus that has no trouble with the 
events of labour; 
 non-reactive fetal monitor tracing which identify complete loss of reactivity 
and variability which identifies a fetus that is unable to respond [3]. 
Fetal heart rate analysis has become a widely accepted means of monitoring fetal status. 
The most familiar means of acquiring the FHR is Doppler ultrasound [2]. 
FHR monitoring using the Doppler shift resulting from the movements of the heart is a 
standard examination in most obstetrical wards. It consists in aiming an ultrasonic beam 
at the fetal heart. The ultrasound reflected from the heart walls and/or valves is slightly 
Doppler shifted as a result of the movements of the heart. After demodulation this 
Doppler shift is used to extract the FHR and the results plotted continuously as a 
function of time. Most fetal monitors use continuous wave ultrasound but some of the 
latest models use pulsed ultrasound. The ultrasonic frequencies used range from 1-2 
MHz [4]. 
The FHR variability is related to the autonomic nervous system and it is an important 
parameter to assess fetal distress. 
In addition, FHR monitoring is also carried out using the fetal Magnetocardiogram 
(fMCG) that uses superconducting quantum interference device magnetometers. Apart 
from this, fetal Phonocardiography (fPCG) allows the heart sounds to be detected for 
FHR monitoring [5] [2]. 
The Cardiotocography (CTG) is a technical means of recording the fetal heartbeat and 
the uterine contractions during pregnancy. Mainly, these recordings are made by two 
separate transducers, one for the measurement of the fetal heart rate and a second one 
for the uterine contractions. A typical CTG reading is printed on paper and/or stored on
Fetal Monitoring 
 
5 
 
a computer for later reference by the physician. CTG is being used to identify signs of 
fetal distress. 
The majority of FHR analysis techniques are performed using a bedside monitor over a 
relatively short period, with the mother to be in a recumbent position. All of the above 
techniques that are mentioned have been successfully used for FHR monitoring, 
although the initial choice was which of the above techniques would be employed. 
The advantage of the Doppler ultrasound technique is that it can be virtually assured 
that a recording of FHR will be obtained. The disadvantages of such systems are that 
they require intermittent repositioning of the transducer and are only suitable for use 
with highly trained midwifes. 
The ultrasound transducer is 
 A little bit uncomfortable because the women undergoing fetal ultrasounds are 
often advised to drink a certain amount of water about an hour beforehand 
and not to use the bathroom during this time. 
 Problematic because the procedure involves launching a 2 MHz signal towards 
the fetus. The use of Doppler ultrasound (noninvasive manner) is not suitable 
for long periods of FHR monitoring. 
 
Figure 2.2 Use of Doppler Ultrasounds [6]
Fetal Monitoring 
 
6 
 
This may involve skillful placement and continual repositioning of the transducer, 
which would be a severe problem for long-term ambulatory use. It may cause records of 
uncertain accelerations or decelerations and true abrupt changes can be misinterpreted 
as noise. The major limitation of the Doppler ultrasound technique is its sensitivity to 
movement. The movement of the mother can result in Doppler-shifted reflected waves, 
which are stronger than the cardiac signal. Thus the Doppler ultrasound technique is 
inappropriate for long-term monitoring of the FHR, as it requires the patients to be bed-
rested. Moreover, the detection of the heartbeat using Doppler ultrasound relies upon a 
secondary effect (the mechanical movement of the heart) and is therefore not as 
accurate for beat-to-beat analysis as detection of the QRS complex. Allied to this 
drawback is the fact that most Doppler systems rely upon some form of averaging to 
produce their FHR data. 
However, the FHR is the only parameter obtained by Doppler ultrasound, while 
research has shown that a global assessment of morphological and temporal parameters 
of the electrocardiogram (ECG) of the fetus during gestation can provide additional 
information about the fetal well-being [5]. 
2.3 Fetal Electrocardiography 
2.3.1 Electrocardiogram 
The Electrocardiogram (ECG) is a graphical recording of the electrical potentials 
generated in association with heart activity. Aristotle first noted electrical phenomena 
associated with living tissues and Einthoven was able to measure the electrical activity 
of the heart in 1901 that resulted in the birth of electrocardiography [3]. As the heart is 
not directly accessible, cardiac electrical activity is usually inferred from measurements 
recorded at the surface of the body, e.g., at the arms, legs, and chest. 
Electrocardiographic monitoring and diagnosis have become necessary tools in 
operating and recovery rooms. For instance, the ECG is of diagnostic value in the 
following clinical circumstances: 1) atrial and ventricular hypertrophy, 2) myocardial 
ischemia and infarction, 3) pericarditis, 4) systemic diseases that affect the heart, 5) 
determination of the effect of cardiac drugs, especially for digitalis and certain anti-
arrhythmic agents, 6) disturbances in electrolyte balance, especially with respected to 
potassium, and 7) the evaluation of function of cardiac pacemakers.
Fetal Monitoring 
 
7 
 
Cardiac events on an ECG are associated with alphabetical labels as shown in Figure 
2.3. A cycle of normal heart beat, as illustrated in Figure 2.3, consists of waves, 
complexes, intervals, and segments representing as follows: 
 P wave: the firing of the sinoatrial (SA) node and atrial depolarization; 
 PR segment: the delay of the electrical impulse at the atrioventricular (AV) 
node to allow for atrial contraction; 
 PR interval: the atrial depolarization and atrioventricular delay; 
 QRS complex: the ventricular depolarization; 
 Q wave: the initial negative deflection resulting from ventricular depolarization; 
 R wave: the first positive deflection resulting from ventricular depolarization; 
 RR interval: the amount of time between consecutive R waves; 
 S wave: a second negative deflection of ventricular depolarization following the 
R wave; 
 ST segment: a part of the ventricular depolarization process; 
 T wave: the ventricular repolarization; 
 QT interval: the amount of time which ventricular depolarization and 
repolarization take. 
 
Figure 2.3 An example of ECG signal and its components [2]
Fetal Monitoring 
 
8 
 
2.3.1.1 ST Segment: Oxygen deficiency 
Oxygen deficiency is a known cause of neurological damage. Fortunately, intrapartum 
asphyxia with neurological damage or perinatal death is unusual and we have to monitor 
many healthy fetuses to find those that are suffering [7]. The ST interval reflects the 
function of the fetal heart muscle (myocardium) during stress tests. 
During labor, we can assess the condition of the fetus from the only routinely available 
fetal signal, the ECG. 
Intervention according to ST waveform analysis has been found appropriate and it 
results in a significant reduction in the number of acidotic babies. At the same time, 
unnecessary interventions are avoided. 
2.3.1.1.1 ST waveforms 
A horizontal or upward leaning, positive ST segment and a T wave height that is stable 
and does not increase define a normal ST. 
When asphyxia becomes severe and long lasting, the ST waveform returns towards 
normal, in parallel with a markedly reduced ability by the fetus to respond. This also 
means that the same type of change in the ST interval should not be expected as 
asphyxia progresses, simply because the capacity of the fetus to utilize its defenses 
diminishes. 
An increase in T wave amplitude is the classical reaction by a fetus responding to 
hypoxia. A fetus that responds with an adrenalin surge and myocardial anaerobic 
metabolism characterizes this reaction. This pattern signifies that the fetal metabolic 
defense is intact and the fetus thereby has the ability to handle hypoxia. 
A biphasic ST is defined as a downward-leaning ST segment. 
This pattern occurs in two situations. The first is when the fetal heart is exposed to 
hypoxia and has not had the opportunity to respond. The second is when the fetal heart 
has a reduced capacity to respond because it has been exposed to previous stress 
situations and resources are lacking or have already been utilized. Biphasic ST changes 
could also be noted with disturbances in heart muscle function, such as with infections 
or malformations. It appears that the premature myocardium may display more frequent 
biphasic ST events. 
Biphasic STs are divided into three categories; Grade 1 is a downward leaning ST 
segment with the entire segment above the baseline. Grade 2 means that an ST segment
Fetal Monitoring 
 
9 
 
component now crosses the baseline and Grade 3 occurs when the whole of the ST 
segment is below the baseline. 
A significant biphasic event occurs when there are more than two consecutive biphasic 
ECG complexes. With the progression of disturbance in myocardial function, a shift 
from biphasic Grade 1 to Grade 2 and 3 may be seen. 
 
Figure 2.4 Biphasic ST grade [8] 
2.3.1.1.2 Basic definitions 
When we discuss the oxygen deficiency of the fetus during labor, there are three terms 
that need to be distinguished. 
Hypoxemia: a decrease in the oxygen content of the arterial blood alone. 
Hypoxia: a decrease in the oxygen content that affects the peripheral tissues. 
Asphyxia: a general oxygen deficiency that affects the high priority organs as well. 
2.3.1.1.3 Fetal response to hypoxemia 
Hypoxemia is the initial phase of oxygen deficiency. During hypoxemia, the oxygen 
saturation decreases and affects the arterial blood, but cell and organ functions remain 
intact. What we notice is a decrease in oxygen saturation with intact organ function. 
The initial fetal defense against hypoxemia is the more effective uptake of oxygen. 
Reduced activity, in other words a decrease in fetal movements and fetal breathing, may 
serve as another defense mechanism. 
The fetus can handle a situation of controlled hypoxemia for days and weeks. However, 
the development of organ systems may be affected and we should expect a fetus 
exposed to long-term stress to have less ability to handle acute hypoxia during labor.
Fetal Monitoring 
 
10 
 
 
Figure 2.5 Fetal response to Hypoxemia [8] 
2.3.1.1.4 Fetal response to hypoxia 
If oxygen saturation decreases still further, the defense used by the fetus during the 
initial hypoxemia phase may not be sufficient to maintain an energy balance and the 
fetus may then enter the hypoxia phase. This means that the oxygen deficiency now 
starts to affect the peripheral tissues in particular. 
The prime reaction to hypoxia is a fetal alarm reaction with a surge of stress hormones 
and a reduction in peripheral blood flow. This causes the redistribution of blood flow to 
favor the central organs, the heart and the brain. Peripheral tissue anaerobic metabolism 
occurs. 
If hypoxia is limited to the peripheral tissues alone, no fetal damage occurs. The fetus 
can handle this degree of hypoxia for several hours.
Fetal Monitoring 
 
11 
 
 
Figure 2.6 Fetal response to Hypoxia [8] 
2.3.1.1.5 Fetal response to asphyxia 
There is an increased risk of organ failure in connection with asphyxia. The oxygen 
saturation has now become very low and there is a risk of central organ function failure. 
There is anaerobic metabolism in the central high-priority organs and the fetus has to 
utilize its glycogen reserves in the liver and the heart muscle. In the brain, very little 
glycogen is stored and therefore the brain is dependent on glucose supplied from the 
liver. 
 
Figure 2.7 Fetal response to Asphyxia [8]
Fetal Monitoring 
 
12 
 
For this reason, it is very important to obtain a good fECG signal, because ST analysis 
requires good Signal-to-Noise Ratio (SNR). Continuous data are needed to obtain 
reliable ST information [8]. 
2.3.2 The most important Invasive and Noninvasive Techniques  
Electronic fetal monitoring can be external (outside), internal (inside), or both. The 
pregnant woman needs to stay in bed during both types of electronic monitoring, but she 
can move around and find a comfortable position [2]. 
2.3.2.1 Invasive: Internal Monitoring 
Most methods for acquiring the fECG are invasive because internal monitoring involves 
placement of a small plastic device about the size of a pencil eraser through the cervix. 
A spiral wire called the fetal scalp electrode is placed just beneath the skin of the fetal 
scalp. The fetal scalp electrode then transmits direct information about the fetal heart 
rate through a wire to the fetal monitor that prints out this information. Because the 
internal fetal monitor is attached directly to the baby, the fECG signal is sometimes 
much clearer and more consistent than with an external monitoring device. However, 
there may be a slight risk of infection with internal monitoring. 
Obviously, a fetal scalp electrode cannot be used antepartum period as there is a 
significant risk of causing a mark or small cut on the fetal head; the instrumentation 
required for the acquisition of the fMCG is also too cumbersome for ambulatory use [9] 
[2]. 
2.3.2.2 Noninvasive: External Monitoring (AECG) 
In contrast, methods utilizing the Abdominal Electrocardiogram (AECG) have a greater 
prospect for long-term monitoring of FHR (e.g., 24 h) and fetal well-being using signal-
processing techniques.  
In fact, the fECG is an electrical signal that can be obtained noninvasively by applying 
multi-channel electrodes placed on the abdomen of a pregnant woman; therefore the 
three main characteristics that need to be obtained from the fECG extraction for useful 
diagnosis include [10]: 
 Fetal heart rate 
 Waveform amplitudes
Fetal Monitoring 
 
13 
 
 Waveform duration 
The detection of fECG signals with powerful and advanced methodologies is becoming 
a very important requirement in biomedical engineering with the increasing in fECG 
signal analysis in clinical diagnosis and biomedical applications. The fECG contains 
potentially valuable information that could assist clinicians in making more appropriate 
and timely decisions during labor, but the fECG signal is vulnerable to noise, and 
difficulty of processing it accurately without significant distortion has impeded its use. 
Unobtrusive, long-term and risk-free monitoring are obviously the major requirements 
for obtaining the fECG. However, there is not yet a good stated and reliable method for 
fECG extraction. 
The major disadvantages with this technique are that the acquisition of the fECG cannot 
be guaranteed and often has a very low SNR because of the interference caused by 
mECG. In addition, at around the 28th to 32nd weeks of gestation, the amplitude of the 
fECG is markedly attenuated due to the electrically insulating effect of the vernix 
caseosa covering the fetus and the existence of preferred conduction pathways between 
the fetal heart and maternal abdomen around this time [2]. The Electromyography 
(EMG) signal due to an uncomfortable position of the patient or to the uterus 
contraction, together with motion artifact lead to difficulties in determining the FHR 
from the AECG signal. 
To overcome the above problems, some multiple-lead algorithms use the thoracic 
mECG to cancel the abdominal mECG, though this is inconvenient for the patient 
during long-term monitoring. Hence, to make the AECG suitable for the detection of the 
fECG, the SNR must be enhanced. The decision was therefore made to base the 
investigation on the possibility of constructing an ambulatory FHR recorder around the 
acquisition of the abdominal fECG. 
2.3.3 Interfaces and Noise affecting the fECG Signal 
The fECG exhibits a bandwidth of 0.05-100 Hz. In an abdominal recording, the 
maximum amplitude of the QRS usually oscillates from 100 to 150 µV for the maternal 
recording and up to 60 µV for the fetal recording. The energy of the latter has been 
estimated to be less than one quarter of the total signal energy. The fECG signals are 
often obscured by electrical noise from other sources. Common ECG noise sources,
Fetal Monitoring 
 
14 
 
such as power line interference, muscle contractions, respiration, skin resistance 
interference, and instrumental noise, in addition to EMG and Electrohysterogram 
(EHG) due to uterine contractions, can corrupt fECG signals significantly. The shape 
and structure of the fECG signal also depend on the placement of the electrodes (there is 
no standard electrode positioning for optimal fECG acquisition), the gestational age, 
and the position of the fetus. All of the aforementioned constraints make the fECG 
detection and extraction a difficult process. Therefore, it is important to understand the 
characteristics of the electrical noise. 
Electrical noise, which will affect fECG signals, can be categorized into the following 
types [2]: 
1. mECG signal: mECG is the most predominant interfering signal with fECG in the 
abdominal signal. The frequency spectrum of this noise source partially 
overlaps that of the ECG and therefore filtering alone is not sufficient to 
achieve adequate noise reduction. 
 
2. Power line interference: power line interference consists of 50-60 Hz pickup 
and harmonics, which can be modeled as sinusoids and combination of 
sinusoids. Characteristics that might need to be varied in a model of power line 
noise include the amplitude and frequency content of the signal. 
These characteristics are generally consistent for a given measurement 
situation and once set, will not change during a detector evaluation. Due to the 
power line noise, the peak-to-peak amplitude caused by mains frequencies can 
reach up to 50% of peak-to-peak ECG amplitude. 
 
3. Maternal muscle noise: muscle noise is due to maternal movement, often from 
the leg and abdominal muscles and may be picked up from the reference pad 
on the maternal thigh. 
EMG activity in the muscles of the abdomen and uterus is the source of this 
kind of noise. Sometimes, it is difficult to identify the EMG signal in the 
abdominal signal.