DETECTION OF WAVE V USING CONTINUOUS WAVELET TRANSFORM
AND INSTANSTANEOUS ENERGY FOR HEARING LOSS
MOHD RUSHAIDIN BIN MUHAMED
UNIVERSITI TEKNOLOGI MALAYSIA
DETECTION OF WAVE V USING CONTINUOUS WAVELET TRANSFORM
AND INSTANSTANEOUS ENERGY FOR HEARING LOSS
MOHD RUSHAIDIN BIN MUHAMED
A thesis submitted in fulfilment of the
requirements for the award of the degree of
Master of Engineering (Electrical)
Faculty of Electrical Engineering
Universiti Teknologi Malaysia
FEBRUARY 2011
iii
Praise to Allah the Almighty
Thanks to my beloved mother, father and lovely wife
For humanity, hope to be more submitted, pious and to the Creator do we return
iv
ACKNOWLEDGEMENT
Alhamdulillah, thanks to Allah, God the Almighty for giving me the strength and
time to complete this project.
In preparing this thesis, I was in contact with many people – researchers,
academicians, and practitioners. I would like take this opportunity to thank them.
Particular, I would like to express my gratitude to my thesis supervisor, Prof. Ir. Dr.
Sheikh Hussain Shaikh Salleh for his encouragement, guidance, critics, advice and
motivation in the process of preparing and completing this project. I also wish to express
my sincere appreciation to Prof. Dr. Dr. Daniel J. Strauss, Dr. Delb and Farah Corona
for their particular guidance. Without the continued support and interest of these
individuals, this thesis would not have been as complete as what you can see in front of
you.
In particular appreciation is due to En. Najeb Jamaludin, En. Zamri Mohd Zin,
En. Yusman Nawazir and others who have provide assistance at various occasions. Their
views were useful indeed in completing this project. In addition, there were Ear, Nose &
Throat (ENT) Specialists and Audiologists who have assisted me in providing
information in this research. Unfortunately, it is not possible to list all of them in this
limited space. The information they contributed have enriched my understanding and
thoughts.
v
ABSTRACT
Hearing screening is an essential test to detect hearing ability of someone. A
good hearing ability is absolutely necessary for a normal speech development. In cases
with pronounced hearing loss even no speech ability will develop at all causing serious
communication problems and impaired intellectual and emotional development.
Auditory brainstem response is a well known method that has been used on detection of
hearing problem. An alternative method was also introduced in this thesis in detecting
the hearing loss problem. Evoked potentials are used because they can indicate problem
along nerve pathways that are too subtle to show up during neurologic examination or to
be noticed by the person. This description may not be visible on magnetic resonance
imaging test. This study proposes methods of identifying hearing loss based on
continuous wavelet transforms (CWT) and instantaneous energy (IE). Study showed that
the IE performed better than the CWT with the sensitivity of 0.88 and specificity of 0.81
compared to the CWT with the sensitivity of 0.63 and specificity of 0.83. Significance of
the results is 0.42.This experiments result can be used as a basis to improve methods of
detection hearing loss.
vi
ABSTRAK
Ujian pendengaran merupakan satu ujian yang penting bagi mengesan keupayaan
pendengaran seseorang. Keupayaan pendengaran yang baik sangat perlu untuk
pembangunan pertuturan yang normal. Dalam kes - kes kehilangan pendengaran,
keupayaan pertuturan juga tidak dapat dibangunkan yang mana menyebabkan masalah
komunikasi yang serius dan juga ketidakseimbangan pembangunan intelektual dan
emosi. Auditori refleks batang otak merupakan kaedah yang sering kali digunakan bagi
mengesan masalah pendengaran. Keupayaan rangsang digunakan kerana ia boleh
mengesan masalah sepanjang lorongan saraf yang terlalu halus untuk dilihat semasa
pemeriksaan neurologi atau dikesan oleh seseorang dengan mata kasar. Masalah
sepanjang lorongan saraf ini mungkin tidak dapat dilihat melalui pengimejan resonans
magnetik. Dalam kajian ini kaedah mengenalpasti gangguan pendengaran berdasarkan
transformasi gelombang selanjar (CWT) dan tenaga serta merta (IE) telah dicadangkan.
Penyelidikan menunjukkan bahawa IE lebih baik daripada CWT dengan tahap kepekaan
0.88 dan kespesifikan 0.81 berbanding CWT dengan tahap kepekaan 0.63 dan
kespesifikan 0.83. Keputusan ekperimen ini boleh digunakan sebagai asas untuk
meningkatkan kaedah pengesanan kehilangan pendengaran.
vii
TABLE OF CONTENTS
CHAPTER
1
TITLE
PAGE
DECLARATION
ii
DEDICATION
iii
ACKNOWLEDGEMENT
iv
ABSTRACT
v
ABSTRAK
vi
TABLE OF CONTENTS
vii
LIST OF TABLES
x
LIST OF FIGURES
xi
LIST OF SYMBOLS
xiii
LIST OF ABBREVIATIONS
xv
LIST OF APPENDICES
xvii
INTRODUCTION
1.1
Introduction
1
1.2
Problem statement
2
1.3
Objectives of the Research
3
1.4
Scope of Research
4
1.5
Contribution of the Thesis
5
1.6
Outline of Thesis
5
viii
2
LITERATURE REVIEW
2.1
Introduction
7
2.2
Hearing Loss
8
2.2.1
Types of Hearing Loss
9
2.2.2
Hearing Loss Factors
9
2.3
2.4
2.5
3
10
2.3.1
10
Hearing Screening Tools
Digital Signal Processing Technique
13
2.4.1
Fourier Transform
13
2.4.2
Averaging Technique
14
2.4.3
Instantaneous Energy
16
2.4.4
Wavelet Transform
17
Summary
19
HEARING SYSTEM
3.1
Introduction
21
3.2
Human Ear
22
3.2.1
Outer Ear
22
3.2.2
Middle Ear
22
3.2.3
Inner Ear
23
3.3
Auditory Nervous System
27
3.4
Human Brain
28
3.4.1
28
3.5
3.6
4
Hearing Screening
Brainstem
Evoked Potential
30
3.5.1
Electroencephalography
32
3.5.2
Auditory Brainstem Response
34
Summary
37
DATA COLLECTION
4.1
Introduction
38
4.2
ABR Recording
38
4.2.1
Subject and Environment Description
39
4.2.2
Electrodes
39
ix
5
7
43
4.2.4
Hardware
45
Data acquisition
48
4.4
Summary
49
DATA ANALYSIS
5.1
Introduction
50
5.2
Averaging Technique
50
5.3
Continuous Wavelet Technique
52
5.3.1
Preprocessing
54
5.3.2
Wave V Detection
55
Instantaneous Energy Technique
57
5.4.1
Preprocessing
58
5.4.2
Wave V Detection
58
5.5
Sensitivity and Specificity
59
5.6
Summary
61
RESULT AND DISCUSSION
6.1
Introduction
62
6.2
Averaging Technique
62
6.3
CWT Technique
69
6.4
IE Technique
75
6.5
Performance Analysis
80
6.6
Summary
82
CONCLUSIONS AND RECOMMENDATIONS
7.1
Conclusions
83
7.2
Recommendations
84
REFERENCES
Appendices
Stimulus
4.3
5.4
6
4.2.3
87
97-101
x
LIST OF TABLES
TABLE NO.
TITLE
PAGE
2.1
Hearing loss ranking and its corresponding decibel
ranges
8
2.2
Screening tools versus age level
12
4.1
Electrodes configurations
40
4.2
Electrodes configuration used by other researchers
41
4.3
Parameter setting that had been used by other
researchers
44
5.1
Mneumonic device for calculate sensitivity and
specificity
60
61
CWT experiment results on hearing loss persons.
73
6.2.1
CWT experiment results on normal persons
74
6.2.2
CWT experiment results on normal persons
75
6.3.1
IE experiment results on normal persons
78
6.3.2
IE experiment results on normal persons
79
6.4
IE experiment results on hearing loss persons
79
6.5
Sensitivity and specificity of CWT technique
81
6.6
Sensitivity and specificity of IE technique
81
xi
LIST OF FIGURES
FIGURE NO.
TITLE
PAGE
3.1
Anatomy of the human ear. From Wikipedia site
24
3.2
Details of ossicles. From webschoolsolutions site
25
3.3
Cochlea-cross sections. From Wikipedia site
25
3.4
Details of human ear anatomy. From Park Avenue
Acoustics site
26
3.5
Human brain structure .(Noback, The Human
Nervous System, 4th ed., McGraw-Hill, 1991)
29
3.6
Typical set of EEG signals (Adopted from Sanei
and Chambers, 2007)
33
3.7
Human auditory pathways. (Netter,2003)
34
3.8
Human auditory pathways and the correlation with
the ABR signal (Courtesy of Grass Telefactor, an
Astro-Med. Inc. Product Group, West Warwick,
RI.)
35
4.1
Electrodes that were used in the experiment
39
4.2
Types of recording (a) ipsilateral (b) contralateral
40
4.3
Electrodes positions on the scalp
42
4.4
Stimulus (a) click (b) tone burst (c) chirp
43
4.5
Trigger box
45
xii
4.6
gUSBamp biosignal amplifier
46
4.7
Programmable attenuator
47
4.8
Hardware configuration
47
4.9
Matlab model configuration
48
5.1
Signal segmentation
52
5.2
Time and frequency resolution cells for wavelet
54
5.3
Continuous wavelet of ABR signal
57
5.4
Color code of CWT contour
57
5.5
Instantaneous Energy of ABR signal
59
6.1
Power spectrum of raw signal
63
6.2
Comparison between raw signal and averaged
signal of 2000 sweeps
63
6.3
Power Spectrum; (a) 1000 (b) 800 (c) 625 (d) 500
sweeps
64
6.4
Power Spectrum; (a) 250 (b) 125 (c) 100 (d) 80 (e)
50 (f) 40 (g) 20 (h) 10 sweeps
65
6.5
The ABR averaged signal and its power spectrum
66
6.6
ABR waveform
68
6.7
ABR waveform of various sweeps
68
6.8
(a) Averaged signal (b) CWT of 1000 sweeps
averaged signal
70
6.9
(a) Averaged signal (b) CWT of 10 sweeps
averaged signal
71
6.10
Abnormal; (a) Averaged signal (b) CWT of 10
sweeps averaged signal
72
6.11
Normal: (a) Averaged signal (b) IE of 1000 sweeps
averaged signal
76
6.12
Abnormal: (a) Averaged signal (b) IE of 1000
sweeps averaged signal
77
xiii
LIST OF SYMBOLS
dBnHL
-
Decibel normal hearing level
μV
-
microvolt
Ag-AgCl
-
Argentum-argentum cloride
k
-
kilo
:
-
Ohm
A1
-
Mastoid left
A2
-
Mastoid right
Cz
-
Vertex
Hz
-
Hertz
Fz
-
Forehead
dB
-
Decibel
μs
-
microsecond
dB SL
-
Decibel sensation level
dB peSPL
-
Decibel peak equivalent sound pressure level
ms
-
millisecond
s
-
second
-
translation
-
wavelet transform
-
integral
-
Square root
xiii
t
-
time
-
scale
-
summation
xv
LIST OF ABBREVIATIONS
ABR
-
Auditory brainstem response
AC
-
Alternating current
AEP
-
Auditory evoked potential
ASSR
-
Auditory steady state responses
BAEP
-
Brainstem auditory evoked potentials
CD
-
Compact disk
CNR
-
Contrast to noise ratio
CT
-
Computed tomography
CWT
-
Continuous wavelet transform
DPOAE
-
Distortion product otoacoustic emissions
DSP
Digital signal processing
DWT
-
Discrete wavelet transform
ECG
-
Electrocardiography
EEG
-
Electroencephalography
EMG
-
Electromyography
EMI
Electro-magnetic interferences
EOAE
-
Evoked otoacoustic emissions
EP
-
Evoked potentials
ERP
-
Event related potential
FFT
-
Fast Fourier transform
xv
gPAH
-
Guger programmable attenuator headphone buffer
gTec
-
Guger Technology
gUSBamp
-
Guger USB amplifier
g.Zcheck
-
Guger impedance check
ICU
-
Intensive care units
IE
-
Instantaneous energy
LED
-
Light emmiting diode
MRI
-
Magnetic resonance imaging
NICU
-
Neonatal intensive care units
OAE
-
Otoacoustic emissions
OAFCCD
-
Ontario association for families of children with
communication disorders
OR
-
Operating rooms
SEP
-
Somatosensory evoked potentials
TEOAE
-
Transient evoked otoacoustic emissions
USB
-
Universal serial bus
UTM
-
Universiti Teknologi Malaysia
VEP
-
Visual evoked potentials
WACFM
-
Weighted averaging based on criterion function
minimization
CHAPTER 1
INTRODUCTION
1.1
Introduction
Hearing screening is an essential test to detect hearing ability of someone. It
is one of the most important recommendations in modern pediatric audiology. A
good hearing ability is absolutely necessary for a normal speech development. In
cases with pronounced hearing loss even no speech ability will develop at all causing
serious communication problems and impaired intellectual and emotional
development. Many methods have been used as hearing screening tools. One of the
popular screening tools is auditory brainstem response (ABR) machine (Satoshi et
al., 2003; Bradley and Wilson, 2004). There are also some other hearing screening
tools such as transient evoked otoacoustic emissions (TEOAE), distortion product
otoacoustic emissions (DPOAE), pure tone audiometry, tympanometry and auditory
steady state responses (ASSR) (Satoshi et al., 2003; McWilliams, 2008; Swanepoel
et al., 2004).
Nowadays, hearing screening becomes important. There are needs to
implement hearing screening in every hospital. However, the time consuming
problem for the existing equipment have caused difficulties to implement hearing
2
screening. Recently, many researchers try to come out with new approach to improve
the hearing screening system.
1.2
Problem Statement
Several types of hearing screening programs have been established so far to
detect the hearing loss as early as possible (Helfand et al., 2001; Delb, 2002; Delb,
2003). The technical methods used in these programs include otoacoustic emissions
(OAE, sound responses that are emitted from the ear) and auditory evoked responses
or ABR (responses in the electroencephalography that are evoked by an auditory
stimulus).
There are two types of OAE screening techniques, TEOAE and DPOAE
(Helfand et al. 2001; Delb, 2003; Plinkert and Delb, 2001; Delb et al., 1999; Delb et
al., 2004). TEOAE are generated in response to clicks while DPOAE are a response
to tones. Both stimuli are presented via lightweight ear canal probes. A microphone
picks up the signal, and multiple responses are averaged to get a reproducible
waveform. This test can be carried out at the bedside and a pass or fail response is
recorded. TEOAE measurements are more commonly used for screening whereas
DPOAE are still a subject of research (Delb et al., 1999). The absence of TEOAE
indicates that the inner ear is not responding appropriately to sound. TEOAE can be
used for a hearing check but they do not allow for a quantification of the hearing loss
(Delb, 2003).
The ABR is an electrophysiological response in the electroencephalography
generated in the brainstem in response to auditory signals such as clicks or chirps.
The stimulus is delivered via earphones or an inserted ear probe, and scalp electrodes
are used to obtain the signals. Detection of wave V in the ABR measurements is the
3
most reliable objective diagnosis and quantification of hearing loss (Delb, 2003;
Wicke et al., 1978; Woodworth et al., 1983; Mason and Adams, 1984; Peters, 1986;
Shangkai and Loew, 1986). This method has a higher specificity as the TEOAE
measurement and it can also be used for the detection of the hearing threshold, i.e.,
the quantification of the hearing loss.
However, due to a poor signal-to-noise ratio, 2000 to 4000 sweeps have to be
averaged to obtain a meaningful, visually noticeable signal at a particular stimulation
level (the exact number depends on the number of artifacts produced). As such largescale averaged signals are used in the conventional visual analysis; they are also
commonly used in computational scheme, although, for a machine other data
representations might be more appropriate (Strauss et al., 2004). Using the currently
available devices, this takes approximately 2 to 4 minutes to get the result for one
stimulation level, e.g., (Meier et al., 2004) where it was not possible to obtain a
reliable response in less than even 4 to 5 minutes. This measurement time requires a
state of spontaneous sleep or strong sedation. Therefore, the needs to design a faster
screening tool become important.
1.3
Objectives of the Research
The main objective of this research is to use signal processing theory to
improve the detection of wave V with reduced number of sweeps. In order to
achieve the main objective, several sub-objectives are addressed in this thesis as
following:
(1) To apply averaging technique in detecting the wave V signal for
abnormalities of hearing impairment.
4
(2) To apply Fourier transforms to analyze the power spectrum on
different sweeps or clicks.
(3) To apply continuous wavelet transforms to detect abnormalities of
hearing impairment.
(4) To apply instantaneous energy to detect abnormalities of hearing
impairment.
1.4
Scope of Research
The scopes of research are limited to the following issues:
x
The research is focus on real ABR signal. The simulated signal is not
included in this research. Real ABR signal is original ABR signal recorded
from patient, while simulated signal is generated by machine known as ABR
signal simulator. Analysis and features extraction based on simulated signal
are not reliable. Real signal are largely interfered by many factors such as
background noise, patients’ movements and age.
x
The real ABR signals are collected at out-patient clinic at Pusat Kesihatan
Universiti Teknologi Malaysia (UTM) and Pusat Kejuruteraan Bioperubatan,
Fakulti Bioperubatan dan Sains Kesihatan, UTM.
x
The patients’ ages are from eight to thirty years old.
5
x
This research is limited to data analysis based on ipsilateral recording.
x
This research focused on the averaging technique, continuous wavelet
transforms (CWT) and instantaneous energy (IE) to detect abnormalities of
hearing impairment and this research used Fourier transform to analyze the
power spectrum on different sweeps or clicks.
1.5
Contribution of the Thesis
In this research, there are two different techniques have been developed based
on signal analysis techniques which are IE and CWT used as marker in order to
detect wave V. The main contributions of this research is to see whether signal
processing theory can be an alternative approach to determine hearing impairment
compared to the traditional approach of averaging. The CWT and IE approach in this
thesis do provide a primary result to further the work in this area.
1.6
Outline of the Thesis
This thesis is organized into six chapters. The current chapter describes the
introduction of this research, the background and problems of this research as well as
the objectives, scopes and contributions of this research.
Chapter 2 describes literature review on hearing loss and hearing screening.
Various types of hearing loss are described; conductive hearing loss and
6
sensorineural hearing loss. The hearing loss factors are also discussed in this chapter.
Various types of hearing screening tools and the digital signal processing techniques
used on analyzing auditory signals are discussed.
Chapter 3 describes an overview of human ear, auditory nervous system,
human brain and the evoked potential. The human ear from the outer to the inner ear
had been discussed. The discussion also covered sound propagation from the outer
ear to the brain. The evoked potential is also discussed in this chapter.
Chapter 4 describes the methodology of the research in detail. The features of
the subjects and environment are described in this chapter. This chapter covered the
discussion on the electrodes configurations that had been used in this study. The
stimulus, hardware as well as data acquisition are also discussed in this chapter.
Chapter 5 describes the analysis techniques of the research in detail. The most
popular method in ABR research, the averaging technique is discussed in this
chapter. The technique is used to extract the ABR wave signal from the recorded
brain signal. The wave V detection techniques that had been used in this research are
also discussed.
Chapter 6 delivers the results of the experiments carried out. Several
experiments were carried out, based on averaging, continuous wavelet transform and
instantaneous energy techniques to find out a way to establish fast detection of wave
V in order to test hearing problems.
Chapter 7, the final chapter, summaries the research findings and some
suggestions for future work which might be useful for further development and
improvement are written in this chapter.
CHAPTER 2
LITERATURE REVIEW
2.1
Introduction
This chapter describes a literature review of types of hearing loss, hearing
loss factors as well as hearing screening tools. Based on the review of the various
types of hearing screening and current hearing screening system, the limitations of
the hearing screening system are identified: hearing screening stages, time
consuming and no ABR raw signals database. From these problems, the objectives of
this thesis were derived to overcome the above constraints by investigating ways in
developing techniques of detection of wave V signal. The scope of the task is also
defined.
8
2.2
Hearing Loss
Hearing loss, hearing impairment or deafness is defined by Ontario
association for families of children with communication disorders, OAFCCD (2008)
as a full or partial decrease in ability to detect or understand sounds. Deafness is the
traditional term used to describe loss of hearing, but hearing impairment is the
preferred term that encompasses the fact of different degrees of hearing loss. Hearing
impairment persons lost of the ability to detect some frequencies, or to detect lowamplitude sounds. Hearing threshold or hearing sensitivity indicates the quietest
sound that an individual can detect. Normal hearing thresholds are not the same for
all frequencies at the same amplitude of the sounds. A person is considered to have
hearing impairment when he or she is not sensitive to the sounds normally heard. The
severity of a hearing impairment is categorized according to how much louder a
sound must be made over the usual levels before the listener can detect it. In
profound deafness, even the loudest sounds that can be produced by the instrument
used to measure hearing may not be detected.
Severity of hearing loss can be measured by the degree of loudness, as
measured in decibels. The hearing loss quantification can be ranked as mild,
moderate, severe or profound. The details of the ranking of the hearing loss are
shown in Table 2.1 below.
Table 2.1: Hearing loss ranking and its corresponding decibel ranges
Ranking
Corresponding decibel ranges
Mild
Adults: 25 to 40 dB
Children: 20 to 40 dB
Moderate
41 to 55 dB
Moderately severe
56 to 70 dB
Severe
71 to 90 dB
Profound
90 dB or greater
9
2.2.1
Types of Hearing Loss
Typically there are three types of hearing impairment; conductive hearing
loss, sensorineural hearing loss and combination of both. Conductive hearing loss
occurs when sound is not conducted properly through the outer ear, middle ear, or
both. It is generally a mild to moderate impairment, because sound can still be
detected by the inner ear. More severe impairments can occur, particularly in
otosclerosis. Generally, with pure conductive hearing loss, the quality of hearing is
good, as long as the sound is amplified loud enough to be easily heard. Sensorineural
hearing loss is due to insensitivity of the inner ear, the cochlea, or to impairment of
function in the auditory nervous system. It can be mild, moderate, severe, or
profound, to the point of total deafness. The persons with this type of hearing
problem, may be heard at normal thresholds, but the quality of the sound perceived is
so poor that speech cannot be understood. The eardrum of this kind of persons can be
function normally but the vibration of the sounds cannot be transferred to the brain
for the interpretation process.
2.2.2
Hearing Loss Factors
There are several factors which cause hearing loss. For the conductive
hearing loss; it is caused by the ear canal obstruction, middle ear abnormalities, inner
ear abnormalities as well as ostoclerosis (an abnormal growth of bone near the
middle ear). The sensorineural hearing loss is caused by the abnormalities in the hair
cells of the organ of Corti in the cochlea, the VIIIth cranial nerve, the
Vestibulocochlear nerve or the auditory portions of the brain. Most of sensory
hearing loss is due to poor hair cell function. The hair cells may be abnormal at birth,
or damaged during the lifetime of an individual. There are both external causes of
damage, such as noise, trauma and infection, and intrinsic abnormalities, such as
deafness genes.
10
2.3
Hearing Screening
Hearing screening or hearing test provides an evaluation of the sensitivity of
a person's sense of hearing and is most often performed by an audiologist using the
screening tools such as TEOAE, DPOAE, pure tone audiometry, tympanometry,
behavioral audiometry and ABR machine.
2.3.1
Hearing Screening Tools
As mentioned in section 2.3, there are various types of screening tools or
methods in the market. McWilliams (2008) described four types of the screening
tools:
Weber test is considered as one of the most basic tests for detecting hearing
loss. This test will be able to detect two types of hearing loss: unilateral conductive
hearing loss and unilateral sensorineural type. The Weber Test makes use of a tuning
fork. The fork is struck on a surface to produce vibrations. It will then be placed on
top of the media lateral of the skull. A person is said to have a unilateral conductive
hearing loss if one ear hears the sound louder than the other. The ear that hears the
louder sound is the affected one.
Rinne test also uses a tuning fork same as the Weber test. The test compares
how sound is perceived as conducted through the mastoid. A tuning fork is struck to
produce vibrations. The fork stem is then struck on the mastoid of a person. When no
sound can already be heard, the fork is then placed outside the ear. Although the
Webber test and Rinne test have been proven to be effective, these can't be compared
to the hearing test called audiometry.
11
Pure tone audiometry is the formal testing of a person's hearing ability. With
the help of an audiometer, the hearing level of a person may be measured. It may
measure the ability of a person to differentiate between different intensities of sound,
distinguish speech from background sounds, or recognize pitch. In audiometry,
otoacoustic emissions as well as acoustic reflex can also be measured. Results from
audiometry testing can be used to diagnose whether the subject has hearing loss or
other problems with the ear. Unlike the Weber test and the Rinne test, audiometry
testing needs a special soundproof room. It also does not make use of tuning forks.
Instead, it uses a device called the audiometer.
Tympanometry or impedance audiometry is a test usually used to detect
conductive hearing loss. It is also used if nothing apparent is detected through the
Rinne and Weber test. This procedure makes use of an otoscope. This makes sure
that nothing; neither foreign object nor earwax is blocking the path to the eardrum. It
is considered as a foolproof method if ever the findings from the other tests produce
suspiciously inaccurate or anomalous results, and further tests are needed for deeper
hearing level assessment. Tympanometry targets the eardrum's mobility, conduction
of bones, and the condition of the middle ear.
There are also the other types hearing screening tools. Below are the details
of other tools.
Evoked otoacoustic emissions (EOAE) is a test that uses a tiny, flexible plug
that is inserted into the baby's ear. Sounds are sent through the plug. A microphone in
the plug records the otoacoustic emissions (responses) of the normal ear in reaction
to the sounds. There are no emissions in a baby with hearing loss. This test is
painless and is usually completed within a few minutes, while the baby sleeps. There
are two types of EOAE; TEOAE and DPOAE as mentioned in the first chapter.
12
ABR is a test that uses electrodes (wires) attached with adhesive to the
subject scalp. While the subject sleeps, clicking sounds are made through tiny
earphones in the subject's ears. The test measures the brain's activity in response to
the sounds. As in EOAE, this test is painless and takes only a few minutes.
Behavioral audiometry is a screening test used in infants to observe their
behavior in response to certain sounds. Additional testing may be necessary.
Play audiometry is a test that uses an electrical machine to transmit sounds at
different volumes and pitches into your child's ears. Child usually wears some type
of earphones. This test is modified slightly in the toddler age group and made into a
game. The toddler is asked to do something with a toy (i.e., touch a toy, move a toy)
every time the sound is heard. This test relies on the cooperation of the child, which
may not always be given.
Some of the screening tools described above can be used on all ages, but
some are used based on age and level of understanding. Table 2.2 below summaries
it.
Table 2.2: Screening tools versus age level
Age level
Tools
Newborn
Infant
Toddler
Child
Adult
Weber
X
Rinne
X
Pure tone
X
X
Tympanometry
X
X
EOAE
X
X
X
X
X
ABR
X
X
X
X
X
X
X
X
X
X
Behavioral
Play
13
2.4
Digital Signal Processing Techniques
Various types of signal processing techniques are being used by many
researchers for various types of application. In this thesis, Fourier transform,
averaging, instantaneous energy and wavelet transform are described.
2.4.1
Fourier Transform
Fourier transform is a mathematical transform used to transform a time
domain signal into frequency domain. Fourier transform has no concern directly with
time (Morita, 1995), it only concern with frequency. It is a description of how a
signal or an image is constructed from waves, it gives us the amplitude and phases of
all waves that must be added together to generate the original signal (Peters et al.,
1998). Brigham and Morrow (1967) mentioned that the fast Fourier transform (FFT)
is a computer algorithm that computes the discrete Fourier transform much faster
than other algorithms. Many researchers used Fourier transform in various
applications.
Yoganathan et al. (1976) used FFT in the frequency analysis of the second
heart sound in normal man. They analyzed the second heart sound signal and
observed the frequency region of the signal. Licznerski et al. (1998) analyzed
shearing interferograms of tear film using FFT for evaluating tear film stability on
the human eye. The tear film distribution on the cornea is measured by the lateral
shearing interference technique. FFT is applied to consecutive interferograms for
precise and repetitive assessment of the tear film breakup time. The tear breakup time
is evaluated noninvasively by comparing the value of the second momentum of
Fourier spectra calculated from the consecutive interferograms.
14
Mohan and Chui (1987) used FFT in calculating dose distributions for
irregularly shaped fields for three-dimensional treatment planning. The dose
distributions for arbitrarily shaped beams are calculated by two-dimensional
convolution of the relative primary photon fluence distributions and kernels
representing the cross-sectional profiles of a pencil beam at a series of depth.
Convolutions are performed using FFT on an array processor.
Wilson and Aghdasi (1999b) studied the effects of subject’s age, gender and
test ear on FFT results. The study was to investigate the effects and relationships
between subject’s age, gender and test ear at different stimulus intensities, on the
spectral content of the ABR in normal adults. Results show that less frequent
changes in the frequency domain ABR FFT results.
Based on review of Fourier transform, it can be concluded that Fourier
transform can be used in many applications and FFT can be used to shows the ABR
components due to less frequent changes in the frequency domain ABR FFT results.
2.4.2
Averaging Technique
Averaging technique is a signal processing technique applied in time domain,
intended to increase the strength of a signal relative to noise that is obscuring it.
Signal averaging technique is an important technique that allows estimation of small
amplitude signals that are buried in noise (Drongelen, 2007). It is often used to
extract a useful signal embedded in noise. This method is especially useful for
biomedical signals, where the spectra of the signal and noise significantly overlap
(Leski, 2002).
15
Tan et al. (2009) used signal averaging method for noise reduction in
anesthesia monitoring and control. While anesthesia patients’ vital signs such as
anesthesia depth index, blood pressure, heart rate etc. are transmitted through a noisy
wireless channel in a wide area, those transmitted signals will be corrupted by the
transmission noise. It is well understood that within most algorithms that reduce
effects of random noises on signals and systems, some types of signal averaging are
used.
The signal averaging can be used effectively when remote monitoring and
diagnosis are involved. On the other hand, signal averaging introduces dynamic
delays. Such delays will have detrimental effects on closed-loop systems, even
destabilizing the system. Consequently, signal averaging encounters a fundamental
performance limitation in feedback systems. Decaying rate of the averaging window
has significant impact on the performance of the close-loop system. When it is larger
than some value, the close-loop system becomes unstable. A concept of stability
margins against exponential averaging was introduced. Its calculation can be
performed by either the Routh-Hurwitz method or the root-locus method on a
modified system. Furthermore, the strategy for choosing the optimal decaying rate
was derived. The results conclude that fast sampling must be used for improving
noise reduction after optimal filter design. The analysis and design method was
applied to anesthesia patient control problems (Tan et al., 2009).
Gazella et al. (1994) shows that signal averaging can be an effective means of
suppressing motion artifacts on T1-weighted spin-echo magnetic resonance imaging
of the liver at high field strengths. Gazella et al. (1994) compares signal averaging
with phase reordering at 1.5T. Signal averaging resulted in images with significantly
greater liver signal to noise ratio, liver-spleen contrast to noise ratio (CNR), and
liver-lesion CNR than did phase reordering.
16
McCarthy (2007) mentioned that averaging also used in acoustic
environment. The disturbances to transmitted signal such as ambient noise,
reverberation and weather changes can be minimized using averaging technique.
Leski (2002) approved that signal averaging can be formulated as a problem
of minimization of a criterion function. He introduced weighted averaging based on
criterion function minimization (WACFM) and robust
-insensitive WACFM to
establish a connection between weighted signal averaging and robust statistics. The
introduced
of Vapnik’s
-insensitive weighted averaging method is based on weighted version
-insensitive function as a dissimilarity measure. A new method is
introduced as a constrained minimization problem of the criterion function.
Based on the review of averaging technique, it can be concluded that
averaging can be used in many applications and it can be use to eliminate noise from
EEG signal and averaging technique was used in this research.
2.4.3 Instantaneous Energy
Instantaneous energy is basically representing the temporal strength of a time
varying signal (Sh-Hussain, 2008). Instantaneous energy had shown good signal
descriptor in detecting first and second heart sounds using ECG waveform (Malarvili
et al., 2003).
Loutridis (2006), used instantaneous energy density to obtain high values of
energy when defected gear teeth are engaged. The defect of gear teeth can be
detected by the high peak instantaneous energy events. Kong et al. (2008) presented
the method based on variation components-based instantaneous energy for voltage
17
sag source detection. Simulations have been performed to provide the thorough
analysis for system with distributed generation units. The studies showed that the
presented method can effectively detect the location of voltage sag source.
As a conclusion, method of instantaneous energy is selected to analyze
auditory brainstem response in this research. The performance of this technique will
be compared.
2.4.4 Wavelet Transform
Wavelet transform has been proven as a useful tool for signal analysis and it
is widely used in biomedical signal processing and denoising applications (Tan and
Mok, 2008) Wavelet transform provide flexible time-frequency window which
automatically narrow when observing high frequency signal and widens when
observing low frequency signal. These transformations able to overcome the
weakness in short time Fourier transforms which use fix window size (Chui, 1992).
(Tan and Mok, 2008) reported that the translation invariant property of
stationary wavelet transform contribute to the improvement of the denoise signal
quality as well as provide a reference selection guide for Coifman wavelet family and
the number of decomposition level to achieve optimum denoise performance for
ECG signal.
Methods of features extraction and classification of heart sounds based on
wavelet decomposition and neural network are shown by Turkoglu and Arslan
(2001). Daubechies 4 coefficients wavelet is used to decompose the heart sounds into
10 levels. Features were obtained by calculating the average of wavelet coefficients
18
at each level. Using adaptive learning back-propagation algorithm, the neural
network had achieved 96% accuracy out of 100 data in classification of 10 types of
heart sounds.
Lightbody et al. (2006) investigates the use of features extracted from the
wavelet domain to assist in the classification of the ABR waveform. They were
classified the strong responses without error by combining power features from the
time and wavelet domain and applying a negative weighting to test cases where the
presence of an artefact was suspected. The remaining ABR waveforms were passed
to a second stage of classification. Cross-correlation features were extracted from
repeat recordings using wavelet decomposition performed on a moving window of
data within the post stimulus waveform. By separating different frequency levels
within the decomposition a more representative post stimulus section of the
waveform was analyzed. The lower level responses with repeat recordings were
classified to an accuracy of 76.4%.
Li et al. (1995) developed an algorithm based on wavelets for the detection of
QRS, T, and P waves of ECG. The power of waveletes lies in its multiscale
information analysis was used to characterize a signal.
Zhang et al. (2005) compared two approaches between wavelets and
wavelets-bayesian for classification of auditory brainstem response (ABR). Based on
their analysis, the wavelet and waveletormed -bayesian approaches both give better
results for ABRs with higher repetition because a higher signal to noise ratio is
achieved in the ABR with more repetitions. Their finding also shows that the
wavelet-bayesian approach performed better than the wavelet approach.
Lee and Ozdamar (1999) compared the resulting self-organizing features
maps topology obtained with and without wavelet preprocessing. Their finding
19
shows that wavelet preprocessing helped to reduce computational time while
retaining the same promising results.
Maglione et al.(2003) investigated the possibility of estimating wave V of the
auditory brainstem response (ABR) by means of wavelet transform. The performance
of different wavelets basis for approximation of the morphology of ABR’s wave V
was evaluated. They used wavelet functions Meyer, Daubechies 10, Symlet 10, and
Biorthogonal 6.8 on their work, and their finding shows Biorthogonal 6.8 to be the
most appropriate to fulfill the specific requirement. They also analyzed the
performance of the Biorthogonal 6.8 wavelet function to approximate the
morphology of wave V for averaged 1000, 750, 500 and 250 epochs The results are
promising and address the use of the wavelet transform to approximate wave V of
ABR using a few epochs, with the consequent advantage of reducing the total time of
recording.
Based on previous research reported on wavelet transform, it can be
concluded that wavelet transform had shown some significant advantages in
analyzing auditory brainstem response. Thus, there is a need to further investigate the
use of wavelet transform in analyzing auditory brainstem response.
2.5
Summary
This chapter reviews about hearing loss, hearing screening and digital signal
processing techniques. There are several factors that cause hearing loss as discussed
in this chapter. By detecting the cause, the suitable treatment can be applied to the
patients. Based on the reviews, the ABR is a suitable tool to screening hearing loss.
The tool can be used between the age ranges, newborn until adult. However, the
screening tool takes time to be completed and there are two stage of screening have
20
to be gone through. Therefore, this research study had been explored the solutions to
overcome the problems. Signal processing techniques, which are Fourier transform,
averaging, instantaneous energy and wavelet transform were discussed in this
chapter.
CHAPTER 3
HEARING SYSTEM
3.1
Introduction
Speech and hearing are closely related in human communication
development. Human beings develop their speaking ability through listening and
imitating the sound. The speaking ability or skill is acquired since the first day a
baby comes to the world through observation and listening. They will try to mimic
the sound by exercising their mouth muscle and producing the similar sound. Hence,
if the baby has problems in hearing and it is not detected at an early stage, it may
lead to delay of speaking ability development or mute if worst. Consequently,
newborn hearing screening (NHS) has become one of the most important
recommendations in modern pediatric audiology. There are many types of hearing
test equipments that have been used in the hospital, such as TEOAE, DPOAE, pure
tone audiometry, tympanometry, behavioral audiometry and ABR machine as
discussed in the chapter 2. In this chapter the details of the hearing system will be
discussed. The brain and ear anatomy are discussed in this chapter. The evoked
potential produced by the human anatomy system is discussed.
22
3.2
Human Ear
The ear is the sense organ that detects sounds. It is also used for balancing.
There are three parts of the ear; the outer ear, the middle ear and the inner ear. The
outer and middle ear mostly collect and transmit sound.
3.2.1
Outer Ear
The outer ear is the part which is visible and is made of folds of skin and
cartilage. The outer part of the ear that consists of the auricle (pinna) and the external
auditory canal are only concerned with the gathering of sound. It leads into the ear
canal, which is about one inch long in adults and is closed at the inner end by the
eardrum. The human pinna is formed primarily of cartilage without useful muscles,
and it has many small superficial bumps and grooves. The human pinna is unique for
each person; the shape and location of the various bumps and grooves differ
considerably across the population (Yost, 2007). The eardrum is a thin, tough,
circular membrane covered with a thin layer of skin. It vibrates in response to
changes in the air pressure that form sound. The eardrum separates the outer ear from
the middle ear.
3.2.2
Middle Ear
The middle ear is involved in transferring the sound to inner parts of the ear
where the auditory receptors can pick up the sound. The middle ear is a small cavity
which conducts sound to the inner ear by means of three tiny, linked, movable bones
23
called “ossicles”, see Figure 3.2. These are the smallest bones in the human body and
are named for their shape. The hammer (malleus) joins the inside of the eardrum.
The head of the malleus is connected to the next ossicles, the incus (Yost, 2007). The
anvil (incus) has a broad joint with the hammer and a very delicate joint to the stirrup
(stapes). The stapes, which is the smallest bine in the body, consists of the head (or
capitulum), two bony struts called cruca on each side of a flat oval bone called the
footplate. The base of the stirrup fills the oval window which leads to the inner ear.
The tympanic membrane or eardrum is held in place by fibers and cartilage situated
in a bony groove between the outer and middle ears. The tympanic membrane
denotes, anatomically, one boundary of the large cavity known as the middle ear
cavity, or tympanum. The tympanic membrane consists of two sets of fibers; one set
radiates from center to the outside of the membrane and the other set is composed of
rings of fibers. The fibers are very sparse in the upper portions of the tympanic
membrane called the “pars flaccida” (Yost, 2007).
3.2.3
Inner Ear
The inner ear can be divided into three parts; the semicircular canals, the
vestibule, and the cochlea which all are located in the temporal bone region of the
skull (Yost, 2007). The inner ear analyzes sound waves and contains an apparatus
that maintains the body’s balance. The inner ear is a very delicate series of structures
deep within the bones of the skull. It consists of amaze of winding passages, called
the “labyrinth”. When the sound has passed to the cochlea, a snail-shaped organ in
the inner ear, is picked up by the inner hair cells when the basilar membrane
vibrates, see Figure 3.1. In general, the motion of the stapes moves the fluid and
other structures of the inner ear. The motion causes the hair cells of the inner ear to
be stimulated and to elicit neural discharges in the auditory nerve. Thus, the
mechanical energy of sound vibration is changed into neural information within the
inner ear. This process is called mechanical to neural transduction. The inner ear
provides the nervous system with information about the frequency, intensity and the
24
temporal content of acoustic stimulation. Part of the spectral analysis of sound is
provided by the mechanics of the inner ear in a way that can be described as filtering
(Yost, 2007).
Semi circular
canals
Elliptical
window
Incus
Circular
window
Vestibular
nerve
Pinna
Auditory
nerve
Malleus
Cochlea
€
Tympanum
cavity
Auditory
canal
Tympanic
membrane
(eardrum)
Stapes
Eustachian
tube
Figure 3.1: Anatomy of the human ear. From Wikipedia site
The vestibule is the central inner ear cavity. It is bounded on its lateral side
by the oval window, which is located in its wall facing the middle ear cavity
(tympanic wall). The vestibule contains the utricle and the saccule, which are sense
organs of the vestibular system. The cochlea, a small shell-shaped part of the bony
labyrinth contains the primary auditory organ of the inner ear. The cochlea resembles
a tube of decreasing diameter, which is coiled increasingly sharp on it. The basilar
membrane connects to the outer wall of the bony cochlea at the spiral ligament and
completes the division of the canal into two passages (tubes or ducts), except for a
small opening at the apex called the helicotrema. The lower passage of the canal
(scala tympani) has an opening, known as the round window, which is covered with
a thin membrane (round window membrane) that connects back to the tympanic
(middle ear) cavity and separates scala tympani from tympanic cavity. The upper
25
passage of the canal (scala vestibule) is connected to the tympanic cavity via the
footplate of the stapes and oval window and separates scala vestibule from the upper
part of tympanic cavity, see Figure 3.4 (Yost, 2007).
Figure 3.2: Details of ossicles. From webschoolsolutions site
Figure 3.3: Cochlea-cross sections. From Wikipedia site.
26
Crura of stapes
Prominence of lateral semicircular canal
Incus
Malleus
Attic of middle ear (epitympanic
recess)
Facial nerve
Footplate of stapes in oval window
Vestibule
Semicircular canals,
utricle and saccule
Internal acoustic meatus
Cochlear nerve
Vestibular nerve
Facial nerve
Pinna
Pharynx
External auditory
meatus (ear canal)
Scala
vestibuli
Ear drum (tympanic
membrane)
Cavity of middle ear
Promontry
Cochlear duct
containing
organ of corti
Scala tympani
Cochlea
Round window
Eustachian tube
Figure 3.4: Details of human ear anatomy. From Park Avenue Acoustics site.
27
The basilar membrane vibrates at the frequency of the stimulus sound and the
vibration rate is then translated into an appropriate number of impulses per second.
The basilar membrane located within the cochlea of the inner ear. It is a stiff
structural element that separates two liquid-filled tubes that run along the coil of the
cochlea, the scala media and the scala tympani, see Figure 3.3 and Figure 3.4. It was
found that different parts of the basilar membrane respond to different frequencies of
sound. Marieb (2001) mentioned that, the fibers near the oval window (cochlear
base) resonate in response to high frequency pressure waves, and the fibers near the
cochlear apex resonate in time with lower frequency pressure waves.
3.3
Auditory Nervous System
The nervous system is essential to the functioning of the human organism. It
regulates our automatic control systems, integrates and assimilates data from outside
world and our internal organs, and regulates and controls the loco motor system. It
has been compared to a computer with an electrical communications system. It is a
complex interconnection of nervous tissue (Carr and Brown, 2001). The nervous
system has been divided into the somatic and autonomic nervous systems. Each of
these systems consists of components from both the central and peripheral nervous
systems (Enderle et al., 2005).
In auditory system travelling wave vibration of the cochlea causes the
stereocilia of the hair cells to undergo a form of bending (shearing). The shearing of
the stereocilia in turn triggers a neural response in the auditory nerve, and the neural
process of hearing begins. The hair cells and the auditory nerve generate
biomechanical-electrical potentials based on a flow of potassium and sodium into
and out of these neural cells. As a consequence, the intricate motions and interactions
of the motions of the various cochlear structures generate several different types of
electric potentials that are relatively easy to measure (Yost, 2007). The flow of neural
28
information starts in the auditory nerve and then travels to the brainstem and then to
the auditory cortex. The output of the auditory nerve is transferred to the auditory
cortex where it is interpreted.
The human auditory system can detect sounds from 10dB to 120dB without
suffering damage. About 10% of nerve fibers have a dynamic range of 60dB or
more. Human can hear sounds in the frequency range between 20 Hz to 20 kHz.
According to Marieb (2001), our ears are most sensitive to frequencies between 1500
Hz and 4000 Hz.
3.4
Human Brain
Human brain is the center of the human nervous system and is a highly
complex organ. The brain has three main parts, the cerebrum, the cerebellum, and the
brainstem. The brain is divided into regions that control specific functions. Brainstem
lies underneath the cerebrum.
3.4.1
Brainstem
The brainstem consists of the medulla oblongata, pons, midbrain and
diencephalon. It connects the brain with the spinal cord and automatically controls
vital functions such as breathing (Enderle et al., 2005). The medulla automatically
controls heart rate and breathing. Reflex functions such as coughing, sneezing and
vomiting are associated with the medulla. The pons forms a noticeable bulge on the
anterior surface of the brainstem. It functions as a relay station for motor respiratory
29
and auditory fibers from the cerebrum and cerebellum. Other impulses from eye
movement, head muscles and taste sensors also pass through here. The midbrain is a
wedge-shaped portion of the stem. It connects the pons and cerebellum with the
cerebrum and is located at the upper end of the brainstem. Midbrain tissues function
as a motor relay station for fibers passing from the cerebrum to the cord and
cerebellum. Integration of visual and auditory reflexes, including those concerned
with avoiding objects, also occur here. The diencephalon forms the superior part of
the brainstem. As part of the original forebrain, it develops into the thalamus and
hypothalamus (Carr and Brown, 2001). Figure 3.5 illustrated the human brain
structure.
Figure 3.5: Human brain structure.(Noback, The Human Nervous System, 4th
ed., McGraw-Hill, 1991)
The thalamus receives fibers from the hearing structures of the inner ear and visual
system. It also provides pathways for somatic sensory systems. All this sensory
information eventually reaches the cerebrum, where it is processed. The
30
hypothalamus responds to the properties of blood passing through nerve connections.
The endocrine system is controlled through nerve responses, affecting emotional
behavior patterns. Other functions controlled via chemical interaction with the
pituitary gland are temperature regulation, water balance, food intake, gastric
secretion, sexual behavior and sleeping patterns (Carr and Brown, 2001).
3.5
Evoked Potential
An evoked potential or evoked response is an electrical signals generated by
the nervous system of a human or other animal in response to sensory stimuli such as
visual, auditory or somatosensory stimulation. Evoked potential amplitudes tend to
be low, ranging from less than a microvolt to several microvolts. To resolve these
low-amplitude potentials against the background of ongoing electroencephalography
(EEG), electrocardiography (ECG), electromyography (EMG) and other biological
signals and ambient noise, signal averaging is usually required. Signals can be
recorded from cerebral cortex, brain stem, spinal cord and peripheral nerves.
Evoked potential testing are electrophysiologic measurements in which a
stimulus is delivered and its transits through the nervous system is timed. The
commonly used tests include visual evoked response, brainstem auditory evoked
response, and somatosensory evoked response (Dawson, 1998). Evoked potentials
are used to measure the electrical activity in certain areas of the brain and spinal
cord. Electrical activity is produced by stimulation of specific sensory nerve
pathways. Evoked potentials test and record how quickly and completely the nerve
signals reach the brain. Evoked potentials are used because they can indicate
problems along nerve pathways that are too subtle to show up during a neurologic
examination or to be noticed by the person. The disruption may not even be visible
on magnetic resonance imaging (MRI) exam.
31
Somatosensory evoked potentials (SEP) are produced by repetitively
stimulating with brief electrical pulses peripheral nerves and recording at various
locations along the nervous system to the sensory cortex (Bell and Manon-Espaillat,
1998). The somatosensory test examines the sensory system from the peripheral
nerve to the sensory cortex of brain. SEP may be tested in patients with numbness or
weakness of arm or leg, or with suspected lesion in spinal cord or peripheral nerve.
SEP is also common diagnostic test for multiple sclerosis.
Visual evoked potentials (VEP) are electrical potential differences recorded
from the scalp in response to visual stimuli (Deuschl, 1999). VEP is caused by
sensory stimulation of a subject's visual field and is observed using
electroencephalography. Commonly used visual stimuli are flashing lights, or
checkerboards on a video screen that flicker between black on white to white on
black (invert contrast). VEP test examines integrity of visual pathway from retina to
occipital cortex where visual input is perceived in the brain. VEP may be tested for
patient with suspected diagnosis of multiple sclerosis, with complaint of visual
disturbance, or with suspected lesion involving visual pathway.
Brainstem auditory evoked potentials (BAEP) are very small electrical
voltage potentials which are recorded in response to an auditory stimulus from
electrodes placed on the scalp. BAEP reflect neuronal activity in the auditory nerve,
coclear nucleus, superior olive and inferior colliculus of the brainstem. BAEP
typically have a response latency of no more than 6 milliseconds with amplitude of
approximately 1 millivolt. Due to small amplitude 500 or more repetitions of the
auditory stimulus are required in order to average out the random background
electrical activity. Although it is possible to obtain a BAEP to a pure tone stimulus in
the hearing range a more effective auditory stimulus contains a range of frequencies
in the form of a short sharp click. BAEP test examines the integrity of auditory
pathway through the brainstem. BAEP are used mainly for evaluation of auditory
pathway in the brainstem (Misulis and Head, 2003). BAEP may be tested for patients
with hearing problem, dizziness or any lesion involving brainstem. BAEP may also
be examined in patients with suspected diagnosis of multiple sclerosis but it is less
32
helpful for evaluation of multiple sclerosis than VEP and SEP (Misulis,and Head,
2003).
3.5.1
Electroencephalography
Electroencephalography (EEG) is the recording of electrical activity along the
scalp produced by the firing of neurons within the brain. It is a unique and valuable
measure of the brain’s electrical function. It is a graphic display of a difference in
voltages from two sites of brain function recorded over time (Tatum et al., 2008).
Traditionally the EEG is recorded to paper with a chart recorder; however, newer
computer technology has enabled the electrical signals to be digitized and stored on
various devices such as optical disks and digital audiotape (Bell and ManonEspaillat, 1998).
In clinical contexts, EEG refers to the recording of the brain's spontaneous
electrical activity over a short period of time, as recorded from multiple electrodes
placed on the scalp. In neurology, the main diagnostic application of EEG is in the
case of epilepsy, as epileptic activity can create clear abnormalities on a standard
EEG study. A secondary clinical use of EEG is in the diagnosis of coma and
encephalopathy. EEG used to be a first-line method for the diagnosis of tumors,
stroke and other focal brain disorders, but this use has decreased with the
introduction of anatomical imaging techniques such as MRI and computed
tomography (CT).
Derivatives of the EEG technique include evoked potentials (EP), which
involves averaging the EEG activity time-locked to the presentation of a stimulus of
some sort (visual, somatosensory, or auditory). Event-related potentials refer to
averaged EEG responses that are time-locked to more complex processing of stimuli;
33
this technique is used in cognitive science, cognitive psychology, and psychophysiological research. For example, the ABR signal is obtained from averaged EEG
signals after response to several repeated sound stimulus such as click. Figure 3.6
illustrates the typical set of EEG signals.
Figure 3.6: Typical set of EEG signals. (Adopted from Sanei and Chambers, 2007)
34
3.5.2
Auditory Brainstem Response
Figure 3.7: Human auditory pathways.(Netter,2003)
35
Figure 3.8: Human auditory pathways and the correlation with the ABR signal.
(Courtesy of Grass Telefactor, an Astro-Med. Inc. Product Group, West Warwick,
RI.)
Auditory brainstem response (ABR) is a popular method of accessing the
electrical activity of a human or other mammal brainstem. It is an electrical signal
36
evoked from the brainstem of a human or other mammal by the presentation of a
stimulus sound such as a click. It is used as a type of test, usually performed for
infants and young children that evaluate how well sounds travel along the hearing
nerve pathways from the ear to the brainstem. Figure 3.7 and 3.8 show the auditory
pathways of human hearing system and the correlation with the ABR signals.
ABR is a type of auditory evoked potential (AEP) which is a subclass of
event related potential (ERP). ABR signals occur in the first 20 ms after the stimulus
(Bell et al., 2004). It is a short, transient response elicited by short-duration stimuli,
wide-band click or frequency-specific tone burst. Normal ABR has very
characteristic wave morphology, with its prominent wave’s I-V appearing within 110 milliseconds after stimulus, of which wave V has the largest amplitude, see Figure
3.8.
ABR is a combined synchronized response of ON-neurons of the VIIIth nerve
and brainstem, and presents valuable information on their function. It is widely used
in the diagnostics of cochlear vs. retro-cochlear pathology, particularly auditory
neuropathy/auditory dys-synchrony, and intra-operative monitoring (Hall,1992,
Hood,1998, Don and Kwong,2002). In infants and young children, it is widely used
for hearing screening, diagnostics, and predicting hearing thresholds (Hall,1992,
Hood,1998, Sininger, and Cone-Wesson,2002).
The latencies of waves I, III, and V; inter-peak intervals I-III, III-V, and I-V;
and the ratio of amplitudes of waves V and I (V/I ratio) are used in diagnostics, while
identification of wave V is used in screening and threshold prediction. Hence, clear
recording of waveforms is essential for the use of ABR: the clearer the response, the
better the wave morphology, and the easier it is to identify and label the waves to
define their latencies.
37
ABR amplitude is very small, only 0.1-1 microvolt (µV), i.e. less than a
millionth of a Volt. The frequency range of ABR is 50-3,000 Hz (Hall,1992).
Recording such a faint signal requires significant signal amplification and filtering.
Moreover, digital signal processing (DSP) is needed to synchronously collect a large
number of responses (sweeps) to numerous stimuli in order to uncover the response
from the background noise (Sokolov et al., 2005).
3.6
Summary
This chapter discussed on the human ear, starting from the outer ear to the
inner ear. It also discussed about the relationship between the auditory nervous
system, brainstem and ear. The discussion on the EP such as SEP, VEP and AEP are
also done in this chapter. Based on literature review on the characteristic of ABR, it
is found that ABR signal provide valuable information of the ear condition. It shows
that the ABR can be used to detect the hearing problem.
CHAPTER 4
DATA COLLECTION
4.1
Introduction
This chapter describes the materials and methods that have been used in this
study. Subject details such as name, age, subject condition, the environment,
stimulus intensity and electrode configuration were recorded before undergoing the
experiment. The gTec (Guger Technology) equipment was used in this study. The
details are described in sections 4.2.4.
4.2
ABR Recording
The recording of ABR signal was made using the gTec equipment. The
amplitude of the ABR signal is very small, only 0.1-1 microvolt (µV), i.e. less than a
millionth of a Volt (Hall,1992). The recording signal can be displayed on the
computer screen. The recording process will be discussed in the next subsection.
39
4.2.1
Subject and Environment Description
Sixteen subjects in the age range of between 8 to 30 years old were selected
to undergo the recording session. Four of them were having hearing problems.
Subjects were explained the purpose of the recording and have to sign a consent
letter before the recording.
The environment of the recording must be in a quiet condition. The
recordings were done in a quiet room and well air- conditioned so as to create a
comfortable environment for the recording session. The subjects were asked to close
their eyes during recording session. The air-conditioner and handphone were
switched off during recording.
4.2.2 Electrodes
Figure 4.1 in the next page illustrates the electrodes that have been used in the
experiment. There are two types of electrodes in the market that are commonly used,
the Ag-AgCl and Gold electrodes. The Ag-AgCl electrode was being used in the
experiment. Impedance between scalp and electrode must be below 5k:Strauss et
al., 2004).
Figure 4.1: Electrodes that were used in the experiment
40
There are two types of recording; ipsilateral and contralateral. Ipsilateral is a
recording which records the signal from the ear that is given a stimulus. Contralateral
is a recording which records the signal from the other side of the ear which is not
given the stimulus; see Figure 4.2. In this study, the focus is just for ipsilateral
recording. It is because the ipsilateral produce better wave V compared to
contralateral.
Stimulus
Stimulus
(a)
(b)
Figure 4.2: Types of recording (a) ipsilateral (b) contralateral
In Figure 4.2 above, there were three electrodes placed on the head. The
electrodes were the positive electrode which is labeled as channel 1; placed on the
vertex, the negative electrode labeled as reference; placed on the mastoid and the
ground electrode which is placed on the forehead. Beattie at el (1986) shows that this
configurations type is the best to get the good wave V signal. Table 4.1 below shows
the electrodes configuration used in this thesis. Table 4.2 summaries the electrodes
configurations that had been used by other researchers.
Table 4.1: Electrodes configurations
Channel 1 (+)
Reference (-)
Ground
Vertex
Right Mastoid
Forehead
41
Table 4.2: Electrodes configuration used by other researchers
Electrode Placements
NonInverting(+) /
Channel 1
Inverting(-) /
Reference
Common /
Ground
Vertex (Cz)
Mastoid
(A1/A2)
Nasion
Upper
Forehead (Fz)
Mastoid
(A1/A2) – test
ear
Mastoid
(A1/A2) –
non test ear
Boston (1981),
Woodworth et al. (1983)
Vertex (Cz)
Mastoid
(A1/A2)
Upper
Forehead (Fz)
Beltran and Cornejo
(2003), Kaplan and
Ozdamar (1987)
Vertex (Cz)
Mastoid
(A1/A2) – test
ear
Mastoid
(A1/A2) –
non test ear
No.
Researchers
1.
American SpeechLanguage-Hearing
Association(1987), Cherrid
et al. (2005)
2.
Wilson and Aghdasi
(1999a,1999b)
3.
4.
Before the recording can be done, the electrodes have to be placed on the
scalp in order to capture the brain signal. The procedure of placing the electrodes on
the scalp is in six simple steps.
Step 1: The positions on the scalp was identified
The positions of the electrodes are shown in the Figure 4.3 below. The vertex is
labeled as Cz and the upper forehead is labeled as Fz. The ground electrode on the
figure is on the lower forehead position. It is also called as nasion. Fz is sometimes
used to place the positive electrode instead of Cz. For the newborn hearing screening,
Fz is a preferred position. It is because the Cz area is still soft for the newborn.
Mastoid is always labeled as A1 for the left ear and A2 for the right ear.
42
Figure 4.3: Electrodes positions on the scalp
Step 2: The placement area was clean carefully using an abrasive gel to ensure low
impedance for the recording. The gel is has sandy characteristics and is used to
remove stains or dead cells from the scalp.
Step 3: The electrode was then filled with the electrode gel and it is assured there are
no air inclusions to avoid the high electrode impedance.
Step 4: The electrodes then plugged at the desired positions.
Step 5: The impedance of the electrodes against a reference electrode checked using
g.Zcheck. The impedance should be lower than 10k:impedance around 5k :are
optimal.
Step 6: The impedance of an electrode is assured not more than 20k :The steps 2 to 5
was repeated if that happen.. The short distances between the electrodes cables had
been used to minimize inductive and capacitive interferences.
43
4.2.3
Stimulus
Stimulus is an input to a system. In hearing screening, stimulus is a sound
input that delivered to the ear. There are several types of stimulus such as click, tone
burst and chirp. In this study, the click stimulus had been used. Figure 4.4 shows the
example of click, tone burst and chirp stimulus waveform.
Click duration
1 cycle
(a)
(b)
(c)
Figure 4.4: Stimulus (a) click (b) tone burst (c) chirp
44
As mentioned earlier, the click stimulus had been used in this study. The
stimulus parameter had been set up with the duration, 80µs, the repetition rate, 10
clicks per second, and the intensity 80 dB. High rates stimulus decrease the
amplitude of the ABR signal. Low rates stimulus are advisable when a full
complement of ABR wave is necessary. Rates of 10 clicks per second or less are
necessary for maximal definition of all the ABR waves (American Speech Language
Hearing Association, 1987). There are also other stimulus parameter settings that had
been used by other researchers. Table 4.3 summaries the parameter settings that had
been used by other researchers.
Table 4.3: Parameter setting that had been used by other researchers
Stimulus Parameters
No.
Researchers
Duration
(ms)
Type
Rate
(clicks/second)
Intensity
(dB)
1.
Wayne J. and Farzin
Aghdasi
(1999a,1999b)
0.1
Click
21
10-90
dBnHL
2.
H.Sohmer et al.
(1981)
n/a
Click
10 or 20
75
dBnHL
3.
Walker Woodworth
et al. (1983)
0.1
Click
28
65 dB SL
4.
J. Robert Boston
(1981)
0.05
Click
28
65 dB SL
5.
N. Beltran and J.M.
Cornejo (2003)
0.08 or
0.1
Click
28
40, 60,
80
dBnHL
6.
American SpeechLanguage-Hearing
Association(1987)
n/a
Click
11.1, 33.3,
66.6
30-90 dB
peSPL
45
4.2.4
Hardware
The equipment that had been used in this study was developed by Guger
Technologies (gTec). There were about six equipments needed in this study; trigger
box, gUSBamp biosignal amplifier, gPAH programmable attenuator, laptop,
headphone and mp3 player. The laptop, headphone and mp3 player were not gTec
equipments. The details about the gTec equipments will be discussed in this
subsection.
The trigger box is equipment that had been used to generate the trigger signal
from the stimulus input. The trigger signal was used to segment the recorded ABR
signal. The input of the trigger box was connected to the attenuator and the output
was connected to the biosignal amplifier. There are adjustable thresholds that have to
be set. The correct setting of the threshold can be seen by the blinking LED. If the
LED is not blinking, it indicates wrong setting of the threshold. Figure 4.5 below
shows the trigger box that was used in this study.
Figure 4.5: Trigger box
The biosignal amplifier was used to amplify the recorded signal. There are
two types of input signal connected to this amplifier; the brain signal and trigger
signal. The brain signal is connected to the amplifier via electrodes. The trigger
46
signal is connected from the attenuator as mentioned above. The amplifier is a 24 bit
biosignal acquisition device with four independent grounds and free from
interference between the recorded signals. It has an input range of ± 250 mV and it is
supported up to 38.4 kHz sampling rate. In this thesis, the sampling rate used is only
19.2 kHz. The amplifier was connected to the computer or laptop via USB cable. It is
the perfect tool for recording multimodal biosignal data with the highest quality
which allows the investigation of the brain, heart, muscle, eye movement,
respiration, skin response and other body signals. Figure 4.6 below shows the
amplifier that had been discussed.
Figure 4.6: gUSBamp biosignal amplifier
The programmable attenuator was used to attenuate the stimulus input. It is
used to accurately attenuate a sound signal in steps of 1 dB. It can also control the
attenuation using a computer or laptop by utilizing the provided software. The
software is supported to attenuate sound in steps of 10dB. The attenuator is
connected to the computer or laptop via a serial port. The converter serial port to
USB can also be used to connect it. The outputs of the attenuator were connected to
the trigger box and headphone. The input of the attenuator was connected to the mp3
player or CD player or laptop. Figure 4.7 in the next page illustrates the attenuator
that had been discussed.
47
Figure 4.7: Programmable attenuator
The stimulus input was generated using the Matlab software. The generated
file can be burnt to the CD or copied to the mp3 player or just played using a laptop.
The headphone which is connected to the attenuator was used to deliver the stimulus.
The laptop was used to control the attenuation, setting the amplifier configuration,
acquiring the recorded signal using Matlab and to analyze the recorded signal. Figure
4.8 below summaries the connections between the hardware that had been discussed.
Figure 4.8: Hardware configuration
48
4.3
Data Acquisition
Data acquisition is an important process before the analysis can be done. In
this study, the Matlab software version R2006a had been used in order to acquire
data from the biosignal amplifier. All the data had been saved in .MAT format for
analysis using the Matlab. Moreover, the .MAT file has a smaller size compared with
.TXT file. Hence, the file loading process becomes faster.
Figure 4.9 below illustrates the Matlab model configuration that had been
done in this study. The configuration was done in order to link the Matlab software
with the gTec hardware. The biosignal amplifier parameter had been set by using the
g.USBamp block. The amplifier serial number, sampling rate, channel selection and
channel setting can be done in this block. The calibration block is used to calibrate
the amplifier and impedance check block is used to check impedance between
electrode and the scalp.
Figure 4.9: Matlab model configuration
49
4.4
Summary
This chapter had discussed the materials and methods that have been used in
this study. The discussion started with the description of the subject and environment
of the study. The selected electrodes configuration and the procedure to plug the
electrode have also been discussed. This chapter also discussed about the stimulus
and hardware that were used. This chapter ended with the discussion of the
acquisition of data.
CHAPTER 5
DATA ANALYSIS
5.1
Introduction
The analysis techniques that were used in this study are discussed in this
chapter. There are three types of analysis techniques that had been used; averaging
technique, continuous wavelet and instantaneous energy (IE). The analyses of the
data were done offline using Matlab R2006a software.
5.2
Averaging Technique
The averaging technique is the most popular method that has been used to
extract the ABR wave signal from the recorded brain signal. The method is used to
increase the signal to noise ratio in order to reduce background noise of the EEG
signal. There are about 2000 epochs that were needed in order to extract the ABR
signal (Zhang et al.,2004). In the past, there were about 4000 to 8000 epochs used in
this technique.
51
However, the more epochs are used, the more time is needed which then makes the
hearing screening ineffective. The quantity of the epochs needed is dependent upon
several factors. The signal strength and the magnitude of the background noise have
to be considered. The higher the intensity of the stimulus, the less quantity of epochs
is needed (Burkardet al., 2007). The concept of this technique is maintaining the
consistent signal and removing the random signal. The noises are assumed as a
random signal. The recorded EEG signals are assumed to be a sum of ABR and
noise. The EEG signal can be written mathematically as:
x(i) = a(i) + n(i)
(5.1)
and represents one recorded signal, where x(i) is the EEG signal, a(i) is the ABR
signal and n(i) is the noise, which is i is varies from 1 to k which is k is the number
of the sample points.
If m records are collected a matrix of k columns and m rows can be
performed. Therefore, each sampled points can be represented as:
x(ij) = a(ij) + n(ij)
(5.2)
where j goes from 1 to m. Since the ABR signal is identical to each of these records,
the equation 5.2 above can be simplified as:
x(ij) = a(i) + n(ij)
(5.3)
The index j in the noise part cannot be removed because it will differ in each record.
It is because the noise is assumed to be random.
The ABR signal occurs at 1.5 to 20 ms post stimuli (Zhang et al.,2004,
Maglione, 2003). Therefore the window length is always set between 12 to 20 ms. In
this study, 2000 sweeps had been used with the epoch’s time of 20 ms. Figure 5.1 in
the next page shows the segmentation of the recorded signal. The segmentation of
the recorded signal was done using the click stimulus. As discussed in section 4.2.3,
the repetition rate of the stimulus that had been used was 10 clicks per second which
52
means the duration of one click is 100ms, see Figure 5.1. Only the first 20ms at every
single click of the recorded signal was taken out. The other 80ms of the recorded
signal was not being used in this study. It is because the ABR signal only occurs at
the first 20 ms as mentioned earlier. The cut out signals were then averaged out,
which formed the ABR signal.
100ms
20ms
Figure 5.1: Signal segmentation
5.3
Continuous Wavelet Technique
Wavelet transform is one of the mathematical transformations that had been
used to extract information from the raw signal. Wavelet transform is a modified
version of Fourier transform and Short-Time Fourier transform. The wavelet
transform utilizes wavelet functions, giving temporal information on the coefficients.
It is suitable for analyzing EEG signal which consists of low frequency components
for a long duration and high frequency components for a short duration. A wavelet
53
transform maps a time-based signal into functions of scale and time. The scale is
similar to the frequency of the signal and is inversely related to frequency:
(5.4)
A scale basically compresses or dilates the wavelet and hence controls the frequency
component. Large scales correspond to dilated or stretched out signals and small
scale corresponds to compressed signals.
The mathematical representation of the wavelet transformation is:
(5.5)
where
is the wavelet transformation of signal
transforming function and it is called wavelet. The
above is a scale and
and
is the
symbol in the function 5.5
is shift parameter known as translation. Time and frequency
resolution of wavelet depend on the scale, . If
becomes large, the window width in
time domain becomes wider while window width in frequency domain becomes
smaller and vice versa for small . Figure 5.2 illustrates the time and frequency
resolution in wavelet.
54
Frequency
t3
f3
t2
f2
t1
f1
Time
Figure 5.2: Time and frequency resolution cells for wavelet
There are two approaches of wavelet analysis which is continuous wavelet
transform (CWT) and discrete wavelet transform (DWT). The CWT is defined as the
sum over all time of the signal multiplied by scaled and shifted versions of the
wavelet function. In DWT, frequency-domain processing using digital processing
techniques is used to obtain time-scale representation of digital signal. Filters of
different cutoff frequencies are used to analyze the signal at different scales. The
signal is passed through a series of high and low pass filters to analyze the high and
low frequencies, respectively.
5.3.1
Preprocessing
Firstly, the raw data of recorded EEG had been filtered through low-pass and
high-pass filter. Filtering a signal is mathematically convolving the impulse response
of the filter with the signal and can be expressed as:
(5.6)
55
where
is the resultant signal,
is the original signal and
is the impulse
response of the filter. The filter response that had been used was from 100 to 2000
Hz.
The filtered signal underwent the segmentation process as discussed in
section 5.2 above. The segmented signal was positioned in matrix form with 384
columns and about 2000 rows. The quantity of the columns is dependent on the
windows length and the sampling rate. In this thesis, the windows length is 20 ms
and sampling rate is 19.2 kHz. The values of columns got by multiply the windows
length with the sampling rate. The quantity of the rows is dependent on the quantity
of the clicks. The recording time that had been used in this study is about 215
seconds for one recording. It means that the highest number of the rows is 2150. The
rows number can be calculated by multiplying the recording number with the clicks
number per second. The signal was then averaged out.
5.3.2
Wave V Detection
There are many methods of wave V detection that have been introduced by
other researchers. The detection of responses at threshold levels is not trivial and
requires an experienced professional. Different signal detection techniques have been
developed and evaluated to improve test efficiency and reliability (Delgado. and
Ozdamar, 1994). Various approaches like pattern recognition, neural network and
signal processing techniques have been used for automatic detection of wave V
peaks. Habrakan et al. (1993) used neural networks to extract features to identify
peak V in brainstem auditory evoked potential. Wilson et al. (1998 and 1999a) used
discrete wavelet analysis for the peak identification. The wavelets technique is used
to decompose a signal into discrete sets of details (high frequencies) and
approximations (low frequencies). The different scaled signals are then rebuilt from
56
their resulting wavelet coefficients and analyzed in a method similar to the full signal
analysis.
Strauss et al. (2004) introduced an approach to the detection of ABRs using a
smart single sweep analysis system. The method used a small number of sweeps
which is decomposed by optimized tight frames and evaluated by a kernel based
novelty detection machine. Delgado and Ozdamar (1994) mentioned that results of
spectral analysis, spectral filtering and fiber-tract modeling of ABRs were used to
determine the most suitable filters to detect the position of the various peaks. These
analyses revealed general trends in ABR composition from one intensity to another
and were used to write labeling rules. In this study, the CWT and IE of ABR signal
has been introduced as a marker to identify the ABR waves. IE technique has
previously been employed in other applications of ECG and heart sound signal
processing, such as heart sound segmentation (Malarvili et al.,2003). In this section,
the discussion is on CWT technique. The IE technique will be discussed in sections
5.4.
The CWT technique was applied to the signal that had been placed through
the preprocessing as discussed in section 5.3.1 above. The generated CWT of the
signal was then analyzed. The CWT was observed between the time ranges from 4 to
6 milliseconds post stimulus. This is because the study focused on the wave V. There
were some criteria that had been observed in order to analyze the signal. Figure 5.3
below shows the generated CWT which is displayed in contour form. In this study,
the level and the curve of the contour were observed. The level of the contour is
differentiated using the color code as shown in Figure 5.4. The red color is the higher
level and the blue is the lower. The sharp curve with high level contour was marked
using dash line. The line was then had been extended to the averaged signal graph
which is on top of the CWT graph. The details will be discussed in Chapter 6.
57
Frequency (Hz)
Continuous Wavelet (CWT) of 1000 Sweeps Averaged
Times (s)
Figure 5.3: Continuous wavelet of ABR signal
1.5
1
0.5
0
-0.5
-1
-1.5
Figure 5.4: Color code of CWT contour
5.4
Instantaneous Energy Technique
Instantaneous energy (IE) of ABR wave represents the behavior of ABR
wave energy as time progress. IE of a signal can be calculated directly from the
signal amplitude. For signal xn , the complex form, z n of the signal is given by
Sharif et al., (2000):
z n xn jH >xn @
(5.7)
58
where H >xn @ is Hilbert transform of signal xn . The IE E z can be calculated as
follows:
E z n z n z * n (5.8)
where z * n is a complex conjugate of z n 5.4.1
Preprocessing
As discussed in section 5.3.1 the raw data of EEG signal was filtered with the
frequency range between 100 to 2000 Hz. It is then segmented with the window size
of 20 milliseconds before the signal was averaged out in order to extract the ABR
signal. The averaged signal was then processed to generate the IE before the wave
detection can be done.
5.4.2
Wave V Detection
As discussed in section 5.3.2, there are many techniques of wave V detection
that have been introduced. However the quantity of the clicks has been an issue when
it was implemented for hearing screening. The time consuming procedure becomes
an issue too. In this study, the new technique of wave V detection has been
introduced in order to tackle these problems. Many other possible factors have to be
considered in order to tackle the problem. However these studies only focused upon
the issue of wave detection. In this subsection, the IE technique will be discussed.
59
The IE technique that had been discussed in section 5.4 was applied to the
signal after going through the preprocessing. The IE of the signal that had been
generated was then analyzed in order to detect the wave. Figure 5.5 below shows the
example of IE of the ABR signal. There are a few aspects that had been observed in
order to use the IE of the signal as wave detection. The study just focuses upon the
wave V, so only the peak of the IE between the time ranges of 4 to 6 milliseconds
had been observed. The peak was marked using the dash line and had been extended
to the averaged signal graph which is on top of the IE graph. The details of the
analysis will be discussed on the chapter result and discussion.
Instantaneous Energy
IE Representation of Averaged Signal
Time (ms)
Figure 5.5: Instantaneous Energy of ABR signal
5.5
Sensitivity and Specificity
Sensitivity and specificity of a test relate to the ability of the test to identify
correctly both with the disease (sensitivity) and those without the disease
(specificity). Sensitivity is the ratio of the number with the disease who are positive
on the screening test to the number of all those with the disease. In other words,
sensitivity represents the percentage labeled positive on the screening test of all those
who truly have the target condition. Specificity is the ratio of the number of those
60
without the disease who are negative on the screening test to the number of all those
without the disease. In other words, specificity is percentage labeled negative on the
screening test of all those who are truly free of the target condition (American
Speech-Language-Hearing Association, 1997). Table 5.1 below; simplify the
explanation in this paragraph.
Table 5.1: Mneumonic device for calculate sensitivity and specificity
Disease
present
Disease
Absent
Test
positive
True
positives
False
positives
Total
positive
Test
negative
False
negative
True
negatives
Total
negative
Total with
disease
Total without disease
Grand
total
The sensitivity and specificity can be calculated as follows:
Sensitivity and specificity were used to test the performance of CWT and IE
techniques in this thesis.
61
5.6
Summary
This chapter discussed the techniques that had been used in this study.. The
discussion in this chapter started with the explanation of the averaging technique,
continued with the CWT and instantaneous energy techniques that had been used as
wave V detection, and also the preprocessing.
CHAPTER 6
RESULT AND DISCUSSION
6.1
Introduction
This chapter discusses about the results of the wave V detection using the
CWT and the IE techniques and the results of the averaging technique.
6.2
Averaging Technique
The experiment was started with the analysis of various sweeps quantity and
the effect of noise removal. The power spectrum of the filtered and averaged signal
was plotted for several quantities of sweeps. The target of the experiment is to look
at the effect of noise removal by reducing the quantity of the sweeps. Figure 6.1 in
the next page shows the power spectrum of filtered raw signal. The signal was
filtered in the frequency range between 100 to 2000 Hz as mentioned in chapter 5.
There were a lots of background noise in the raw signal as shown in Figure 6.1.
63
Power Spectrum of Raw Signal
0.9
0.8
Normalized Power
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
200
400
600
800
1000
1200
Frequency(Hz)
1400
1600
1800
2000
Figure 6.1: Power spectrum of raw signal
Figure 6.2 shows the comparison between raw signal and the averaged of
2000 sweeps. It is clearly shown that the averaging technique can reduce the
background noise after averaging of 2000 sweeps; see Figure 6.2 (a) and Figure 6.2
(b). The power spectrum of the raw signal and the averaged signal are also different,
see Figure 6.2 (c) and Figure 6.2 (d).
Power Spectrum of Raw Signal
Raw Signal
6
1
0.8
Normalized Power
Amplitude(uV)
4
2
0
-2
-4
-6
0
0.005
0.01
Time(s)
0.015
0.6
0.4
0.2
0.02
0
(a)
500
1000
1500
Frequency(Hz)
2000
(c)
Power Spectrum of Averaged Signal
ABR Averaged Signal of 2000 Sweeps
1
0.8
Normalized Power
Amplitude(uV)
0.5
0
-0.5
-1
0
0.005
0.01
Time(s)
0.015
0.02
0.6
0.4
0.2
0
500
1000
1500
Frequency(Hz)
(b)
(d)
Figure 6.2: Comparison between raw signal and averaged signal of 2000 sweeps
64
The experiment was analyzed on thirteen different quantities of sweeps. The
sweeps were 2000, 1000, 800, 625, 500, 250, 125, 100, 80, 50, 40, 20 and 10 sweeps.
The different spectrum contents between the sweeps had been observed. The number
was chosen because it produced the whole number result. For example; 1000 sweeps
are calculated by dividing the 2000 sweeps by 2 and the 800 sweeps produced by
dividing the 2000 sweeps by 2.5. Figure 6.3 and 6.4 in the next page illustrate the
power spectrum of signal with various sweeps. Figure 6.2 (d), Figure 6.3 and Figure
6.4 (a) show that the power spectrum between 2000, 1000,800, 625, 500 and 250
sweeps are not much different. It shows that there are two high energy of ABR
components in the region 500 and 1kHz (Wilson and Aghdasi, 1999a).
Power Spectrum of 1000 Sweeps Averaged Signal
Power Spectrum of 800 Sweeps Averaged Signal
0.8
Normalize Power
Normalize Power
0.8
0.6
0.4
0.2
0.6
0.4
0.2
500
1000
1500
Frequency(Hz)
2000
500
(a)
2000
(b)
Power Spectrum of 625 Sweeps Averaged Signal
Power Spectrum of 500 Sweeps Averaged Signal
0.8
Normalize Power
0.8
Normalize Power
1000
1500
Frequency(Hz)
0.6
0.4
0.6
0.4
0.2
0.2
500
1000
1500
Frequency(Hz)
(c)
2000
0
500
1000
1500
Frequency(Hz)
2000
(d)
Figure 6.3: Power Spectrum; (a) 1000 (b) 800 (c) 625 (d) 500 sweeps
Power Spectrum of 250 Sweeps Averaged Signal Power Spectrum of 125 Sweeps Averaged Signal
1
1
0.8
Normalize Power
Normalize Power
0.8
65
0.6
0.4
0.2
0.6
0.4
0.2
0
500
1000
1500
Frequency(Hz)
2000
500
(a)
1000
1500
Frequency(Hz)
2000
(b)
Power Spectrum of 100 Sweeps Averaged Signal Power Spectrum of 80 Sweeps Averaged Signal
1
1
0.8
Normalize Power
Normalize Power
0.8
0.6
0.4
0.2
0.6
0.4
0.2
0
500
1000 1500
Frequency(Hz)
2000
0
500
1000
1500
Frequency(Hz)
(d)
(c)
Power Spectrum of 50 Sweeps Averaged Signal
1
Power Spectrum of 40 Sweeps Averaged Signal
1
0.8
Normalize Power
Normalize Power
0.8
0.6
0.4
0.6
0.4
0.2
0.2
0
500
1000 1500
Frequency(Hz)
2000
0
(e)
500
1000
1500
Frequency(Hz)
2000
(f)
Power Spectrum of 20 Sweeps Averaged Signal
1
Power Spectrum of 10 Sweeps Averaged Signal
1
0.8
Normalize Power
0.8
Normalize Power
2000
0.6
0.4
0.2
0.6
0.4
0.2
0
500
1000
1500
Frequency(Hz)
2000
0
(g)
500 1000 1500
Frequency(Hz)
2000
(h)
Figure 6.4: Power Spectrum; (a) 250 (b) 125 (c) 100 (d) 80 (e) 50 (f) 40 (g) 20 (h) 10
sweeps
66
Increase of the power spectrum can be seen in the region 700, 1500 to 2kHz
starting from 125 sweeps to 50 sweeps as shown in Figure 6.4 (b) to (d). The power
spectrum between that frequencies range start decreasing from 40 sweeps to 10
sweeps and the main ABR components cannot clearly seen. We can conclude that,
the more averaged sweeps signal, the better ABR waves we can get. The experiment
then continued by comparing the power spectrum with the averaged signal. The
target of the experiment is to investigate the correlation between the averaged signal
and the power spectrum.
Power Spectrum of 1000 Sweeps
Averaged Signal
Normalize Power
Amplitude (µV)
ABR Averaged Signal of
1000 Sweeps
Frequency (Hz)
(b)
ABR Averaged Signal of
500 Sweeps
Power Spectrum of 500 Sweeps
Averaged Signal
Amplitude (µV)
Normalize Power
Time (s)
(a)
Time (s)
(c)
Frequency (Hz)
(d)
Power Spectrum of 50 Sweeps
Averaged Signal
Amplitude (µV)
Normalize Power
ABR Averaged Signal of
50 Sweeps
Time (s)
(e)
Frequency (Hz)
(f)
Figure 6.5: The ABR averaged signal and its power spectrum
67
Figure 6.5 in the previous page shows the correlation between the averaged
signal and its power spectrum. From the figure, it is shown that the power spectrum
can be used to explain the morphology of the averaged signal. In Figure 6.5 (b), the
high normalize power can be seen in the frequency range between 100 to 200 Hz, the
other frequency contents seem to be low. From those features, the averaged signal in
Figure 6.5 (a) plotted a big ripple, and it can be seen that the small ripple is riding the
big ripple. In Figure 6.5 (d), it is shown that the normalized power of the power
spectrum increased compared to the power spectrum in Figure 6.5 (b). Hence, the big
ripple does not clearly appear in the averaged signal; see Figure 6.5 (c). Figure 6.5 (f)
show an increase of the normalized power between the frequencies ranges 500 to
2000 Hz; therefore it is can be seen the small ripple amplitudes in the Figure 6.5 (e)
are increased. From those phenomena, it is proven that the power spectrum can be
used to explain the morphology of the averaged signal. Besides the comparison
between the averaged signal and its’ power spectrum, the effect of quantity of
sweeps to the averaged signal were also analyzed. Figure 6.5 shows that the quantity
of sweeps is affected by the amplitude of the averaged signal. The amplitudes have
decreased while the quantity of sweeps increased. It is inversely proportional. It is
clearly shown by Figure 6.5 (a), (c) and (e).
The experiment was then continued with the wave V detection using the
averaged signal. Firstly, the detection was done using a single averaged signal. The
amplitude and latency of wave V had been considered in order to mark the wave.
The amplitude is measured between a positive peak and the following negative peak.
By measuring it’s peak to peak, the problem of determining the baseline can be
avoided. Figure 6.6 in the next page shows the example of ABR signal which was
recorded with the click intensity 80 dB. The latency marker that marked in the figure
is just latency of wave V only; the other latencies, such as wave I, II, III and others
are not marked because the study focused on the wave V. The amplitude marker that
marked in the figure shows the measurement range of the amplitude.
68
ABR Averaging Signal
Amplitude (µV)
Latency
Amplitude
Time (s)
Figure 6.6: ABR waveform
Amplitude (µV)
ABR Averaging Signal
Time (s)
Figure 6.7: ABR waveform of various sweeps
69
The detection of wave is much easier by using several averaged signal with
various sweeps by plotting the signals parallel. Figure 6.7 in the previous page
illustrates the ABR waveform of various sweeps from 10 to 2000 sweeps. The
sweeps are plotted in different baseline. The dash line in the figure marks the wave V
of that signal. It can be seen that the peak of the wave consistently appeared in those
signals. However, in some cases the peaks appeared inconsistently especially for the
sweeps below 50. The peaks in the circled area look inconsistent. From the figure, it
can be concluded that the peak of wave V still can be detected while the sweeps
reduced. The next section will discuss the continuous wavelet and IE techniques
which can help to improve detection of the waves.
6.3
CWT Technique
Wavelet transform is a well known technique that had been used in digital
signal processing. As mentioned in section 5.3, there are two approaches of wavelet
analysis which is CWT and DWT. In this section, the discussion focuses on the CWT
technique. The idea of this thesis is to see if we could detect the wave V. Figure 6.8
in the next page illustrates the averaged signal and its CWT representation. The color
code that had been discussed in section 5.3.2 has been used as reference to the
contour colors on the CWT representation. The color code above 0, which is high
level contour, has been used to mark the wave. It can be seen in the figure, three
different lines had been used. The dash line is used to mark the wave V, the other
two lines have been used to mark the border. The wave V in the figure is clearly
marked by using the CWT. Lightbody et al. (2006) had
studied the feature
extraction using DWT. The studies showed that the frequency bands that dominate
the ABR, which are the frequencies in region of 200 Hz, 500Hz and 900 Hz. By
referring to the study, the CWT representations in that region had been observed. . It
has been proved from the experiments that had been done. In Figure 6.8 (b) and
Figure 6.9 (b), it can be seen that the high level contours appeared in that regions.
70
Figure 6.9 (b) shows that CWT can be used to detect ABR waves even though in less
quantity of sweeps.
Amplitude (µV)
ABR Averaging Signal
Time (s)
(a)
Frequency (Hz)
Continuous Wavelet (CWT) of
1000 Sweeps Averaged Signal
Time (s)
(b)
Figure 6.8: (a) Averaged signal (b) CWT of 1000 sweeps averaged signal
71
Amplitude (µV)
ABR Averaging Signal
Time (s)
(a)
Frequency (Hz)
Continuous Wavelet (CWT) of
10 Sweeps Averaged Signal
Time (s)
(b)
Figure 6.9: (a) Averaged signal (b) CWT of 10 sweeps averaged signal
Besides looking at normal results, the study also looked at the abnormal
results, which were collected from hearing loss persons. Initially, the experiments
were done by looking at the CWT representation as a whole and comparing the CWT
between normal and abnormal. It is difficult to differentiate the CWT by looking it as
a whole. Consequently, the experiments continued by looking at certain area from
72
the CWT representations. The area between 4 to 6 milliseconds was used in the
study.
Amplitude (µV)
ABR Averaging Signal
Time (s)
(a)
Frequency (Hz)
Continuous Wavelet (CWT) of 10 Sweeps Averaged Signal
Time (s)
(b)
Figure 6.10: Abnormal; (a) Averaged signal (b) CWT of 10 sweeps averaged signal
Figure 6.10 shows the abnormal of 10 sweeps averaged signal and its CWT.
It can be clearly seen in Figure 6.10 (b) that the contour levels are mostly at the same
level. It is different compared with normal signal, which contains various levels of
73
contours. This event did not appear at every sweep averaged and it does not only
appear when averaging with less quantity of sweeps, it also appeared even in more
quantity of sweeps averaged such as 1000 sweeps. Table 6.1 below summarizes the
recording of abnormal signal of various sweeps on four hearing loss persons with
click intensity of 80 dB. The CWT representations events as discussed in this
paragraph are labeled as NIL for the amplitude, A and latency, L in Table 6.1. It can
be seen that the events occurred on the 10, 20, 40 and 1000 sweeps on three different
persons.
Table 6.1: CWT experiment results on hearing loss persons
No. of sweeps
Subjects
1
2
3
4
1000
A
L (ms)
(uV)
NIL
NIL
0.33
4.38
0.19
4.79
NIL
NIL
Subjects
1
2
3
4
A (uV)
A (uV)
4.32
-0.12
4.38
0.41
5.73
0.12
5.63
0.03
No. of sweeps
500
L (ms)
A (uV)
L (ms)
4.38
4.38
5.73
5.57
-0.25
0.33
0.08
-0.01
4.38
4.43
5.73
5.10
125
L (ms)
4.32
4.43
5.73
5.73
100
A
L (ms)
L (ms)
(uV)
4.32
0.10
4.32
4.48
-0.13
4.48
5.78
1.27
5.78
5.63
0.06
5.05
No. of sweeps
A
(uV)
-0.40
-0.66
0.96
-0.32
50
A
(uV)
-0.50
0.58
3.28
0.10
625
L (ms)
0.07
0.32
0.11
-0.02
250
A
(uV)
-0.05
0.36
0.24
-0.07
Subjects
1
2
3
4
800
40
L (ms)
4.32
4.48
5.78
5.05
A
(uV)
NIL
-0.33
5.69
0.32
80
A (uV)
L (ms)
-0.26
-0.53
1.92
-0.10
4.38
4.48
5.78
5.05
20
L (ms)
NIL
4.48
5.26
5.10
A
(uV)
1.45
NIL
9.77
-0.28
10
L (ms)
A (uV)
L (ms)
4.69
NIL
5.26
5.10
1.07
NIL
19.27
0.05
4.12
NIL
5.26
5.16
74
Table 6.2.1: CWT experiment results on normal persons
No. of sweeps
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
1000
A
L (ms)
(uV)
0.37
5.26
0.30
4.58
0.36
5.63
0.42
4.42
0.46
5.57
0.46
5.73
0.77
5.16
0.59
5.00
0.68
5.00
0.34
5.37
0.52
5.52
0.45
5.05
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
800
A (uV)
L (ms)
0.37
0.44
0.41
0.55
0.47
0.41
0.70
0.51
0.69
0.17
0.50
0.40
250
A
(uV)
0.24
0.84
0.11
1.36
1.06
0.64
1.14
0.64
0.17
0.08
0.68
0.22
625
A (uV)
L (ms)
A (uV)
L (ms)
5.26
4.58
5.57
4.43
5.52
5.78
5.10
5.10
5.00
5.31
5.52
5.05
0.38
0.45
0.31
0.74
0.57
0.58
0.86
0.46
0.58
0.08
0.83
0.32
5.26
4.69
5.57
4.43
5.47
5.78
5.10
5.42
5.00
5.31
5.57
5.05
5.31
0.39
4.64
0.62
5.57
0.37
4.38
0.72
5.52
0.56
5.78
0.50
5.16
0.79
5.10
0.53
5.05
0.62
5.31
0.19
5.52
0.44
5.05
0.39
No. of sweeps
125
500
100
80
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
5.26
4.74
5.73
4.43
5.47
5.83
5.16
5.47
5.10
5.37
5.63
5.05
0.18
0.04
0.32
0.50
1.51
0.61
0.85
0.45
0.40
-0.02
1.45
0.38
5.47
4.84
5.78
4.43
5.52
5.78
5.26
5.73
5.26
5.42
5.73
5.42
0.42
0.27
0.71
0.48
1.40
0.67
1.13
0.32
0.40
-0.08
1.04
0.90
5.47
4.84
5.37
4.38
5.52
5.78
5.78
5.73
5.26
5.42
5.73
5.52
0.56
1.88
0.76
0.64
1.52
0.82
1.15
0.49
0.74
-0.63
1.03
1.59
5.52
4.84
5.37
4.38
5.47
5.83
5.78
5.78
5.26
5.42
5.78
5.47
75
Table 6.2.2: CWT experiment results on normal persons
No. of sweeps
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
50
A
(uV)
0.44
2.29
1.10
0.79
0.36
0.75
0.83
1.07
0.93
-0.51
1.84
0.32
40
20
10
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
5.52
4.29
5.37
4.38
4.95
5.68
5.73
5.83
5.21
5.37
5.78
5.52
0.46
2.02
1.74
1.27
2.44
0.47
3.28
0.38
1.51
-0.61
1.73
0.44
5.47
4.38
5.37
4.32
5.47
5.73
4.64
4.64
5.21
5.37
4.95
4.79
0.42
2.10
2.01
1.54
0.78
1.31
2.78
2.49
1.35
-0.63
2.93
0.48
5.57
4.38
4.90
4.12
5.47
5.68
4.64
4.69
4.74
5.37
4.95
4.74
1.56
5.61
4.43
1.12
NIL
0.15
4.05
NIL
1.02
0.32
3.45
0.54
5.31
3.85
4.38
4.27
NIL
5.73
4.64
NIL
4.84
4.90
4.90
4.79
Table 6.2.1 and 6.2.2 show the CWT experiment results on normal persons.
From the table, it can be seen that the event that occurred on the hearing loss persons,
only occurred at 10 sweeps averaged signal on the normal person, and it involved
only two persons of them. The other feature that differentiate the normal signal with
the hearing loss is the amplitude, A of the signals. It can be seen that the amplitude of
the normal signals are mostly positive in value. Only one person has the negative
amplitude which is starting from 125 sweeps to 10 sweeps. The amplitude of the
hearing loss persons are mostly negative amplitude.
6.4
IE Technique
As discussed in section 5.4, the IE of ABR wave represents the behavior of
ABR wave energy as time progress. Figure 6.11 in the next page shows the
presentation of the ABR averaged signal and the IE of the averaged signal. The
signal was recorded with the click intensity of 80 dB.
76
Amplitude (µV)
ABR Averaging
Time (s)
(a)
Inst Energy
Instantaneous Energy of 1000 Sweeps Averaged Signal
Time (ms)
(b)
Figure 6.11: Normal: (a) Averaged signal (b) IE of 1000 sweeps averaged signal
77
Amplitude (µV)
ABR Averaging Signal
Time (s)
(a)
Inst Energy
Instantaneous Energy of 1000 Sweeps Averaged Signal
Time (ms)
(b)
Figure 6.12: Abnormal: (a) Averaged signal (b) IE of 1000 sweeps averaged signal
The signal in Figure 6.11 (a) and 6.12 (a) are plotted on three different
baselines, 0, 1 and 2 µV, respectively. The averaged signals on the 0 baseline is 1000
sweeps averaged signal, 800 sweeps averaged signal on the baseline 1 and the 625
sweeps on the baseline 2. The vertical dash line marked the latency of wave V. The
line is marked by choosing the peak of IE nearest the wave latency. It can be
78
observed that wave V occurred at specific points in the signal of normal person but
did not occur in the signal of hearing loss person. The peaks that marked by the
vertical line are above the horizontal line in the signal of normal person, see Figure
6.11 (a) and the peaks are below the horizontal line in the signal of hearing loss
person, see Figure 6.12 (a).
Table 6.3.1: IE experiment results on normal persons
No. of sweeps
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
1000
A
(uV)
0.31
0.26
0.35
0.39
0.40
0.46
0.74
0.59
0.64
0.34
0.50
0.43
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
800
L (ms) A (uV)
5.37
4.64
5.57
4.48
4.95
5.73
5.05
5.00
5.10
5.37
5.05
5.00
L (ms)
0.35
0.44
0.41
0.48
0.33
0.40
0.66
0.50
0.67
0.17
0.54
0.36
250
A
(uV)
0.24
0.74
0.10
0.33
0.90
0.50
1.07
0.34
0.17
0.08
0.40
0.22
625
A (uV)
L (ms) A (uV)
5.31
4.64
5.68
4.48
4.90
5.73
5.05
5.00
5.10
5.37
5.10
5.05
0.34
0.37
0.19
0.20
0.15
0.57
0.24
0.52
0.62
0.37
0.23
0.39
L (ms)
A (uV)
L (ms)
5.37
4.64
5.57
4.48
4.95
5.73
5.05
5.00
5.10
5.37
5.05
5.00
0.37
0.42
0.28
0.70
0.21
0.55
0.84
0.42
0.54
0.08
0.52
0.32
5.31
4.64
5.68
4.48
4.90
5.73
5.05
5.00
5.10
5.37
5.10
5.05
5.37
0.34
4.64
0.61
5.57
0.37
4.48
0.70
4.95
0.36
5.73
0.48
5.05
0.78
5.00
0.51
5.10
0.55
5.37
0.18
5.05
0.62
5.00
0.35
No. of sweeps
125
500
100
80
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
5.31
4.64
5.68
4.48
4.90
5.73
5.05
5.00
5.10
5.37
5.10
5.05
0.42
3.18
0.71
0.48
1.40
0.67
1.08
0.59
0.40
0.09
0.94
0.18
5.47
4.17
5.37
4.38
5.52
5.78
4.64
5.00
5.26
4.90
5.00
4.69
0.56
2.12
0.76
0.64
1.52
0.82
1.18
0.50
0.74
-0.12
0.68
0.58
5.47
4.48
5.37
4.38
5.47
5.83
4.64
5.00
5.26
4.79
5.00
4.74
79
Table 6.3.2: IE experiment results on normal persons
No. of sweeps
Subjects
1
2
3
4
5
6
7
8
9
10
11
12
50
A
(uV)
0.44
2.29
1.10
0.79
1.59
0.75
2.09
0.51
0.93
-0.23
1.35
0.43
40
20
10
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
A (uV)
L (ms)
5.52
4.53
5.37
4.38
5.42
5.68
4.64
5.00
5.21
4.84
4.95
4.79
0.46
2.02
1.74
-0.86
2.44
0.47
3.28
0.38
1.51
-0.57
1.73
0.84
5.47
4.38
5.37
5.89
5.47
5.73
4.64
4.64
5.21
4.90
4.95
5.52
0.42
2.10
2.31
-4.38
0.78
1.31
2.78
2.49
0.79
-0.08
2.93
1.24
5.57
4.38
5.37
5.89
5.47
5.68
4.64
4.69
5.21
4.79
4.95
5.73
1.56
2.91
3.96
-6.80
2.06
0.15
4.05
3.39
0.49
0.32
3.45
0.73
5.31
4.38
5.37
5.78
5.52
5.73
4.64
4.74
5.21
4.90
4.90
5.42
Table 6.4: IE experiment results on hearing loss persons
No. of sweeps
Subjects
1
2
3
4
Subjects
1000
A
(uV)
-0.25
0.04
-0.05
-0.06
L
(ms)
5.31
5.31
5.42
5.31
250
A (uV)
1
2
3
4
0.26
-0.99
-0.74
-0.64
Subjects
1
2
3
4
800
L
(ms)
5.37
5.16
5.05
5.42
A (uV)
625
L (ms)
-0.17
-0.22
-0.05
-0.18
A (uV)
5.31
-0.02
5.31
-0.12
5.42
-0.10
5.31
-0.20
No. of sweeps
40
L (ms)
-0.77
-0.46
4.52
0.10
5.21
5.16
5.26
5.05
L
(ms)
5.31
5.31
5.42
5.31
125
100
A
L
A
L (ms)
(uV)
(ms)
(uV)
-0.18
5.37
-0.16
5.21
-0.71
5.16
-0.48
5.10
-0.11
5.05
2.73
5.26
0.84
5.42
0.06
5.05
No. of sweeps
50
A (uV)
500
A
(uV)
-1.29
-1.46
7.69
0.32
A (uV)
L (ms)
0.03
-0.30
-0.40
-0.22
5.37
5.16
5.05
5.42
80
A (uV)
L (ms)
0.03
-0.54
3.64
-0.10
5.26
5.10
5.26
5.05
20
L
(ms)
5.16
5.00
5.26
5.10
A
(uV)
1.08
0.87
9.77
-0.28
10
L
(ms)
5.26
5.00
5.26
5.10
A (uV)
-0.07
-0.31
17.27
0.05
L
(ms)
5.21
5.00
5.26
5.16
80
Table 6.3.1 and 6.3.2 show the IE experiment results on normal persons and
Table 6.4 show the IE experiment results on hearing loss persons. In Table 6.4, it is
shown that the amplitude of the signals for the hearing loss person mostly have
negative amplitude. It means that the waves V do not appear on the hearing loss
persons. For the normal person, two out of twelve have the negative amplitude, see
Table 6.3.1 and 6.3.2. Subject 4 has negative amplitudes starting from 40 sweeps to
10 sweeps averaged signals. Subject 10 has negative amplitudes starting from 80
sweeps to 20 sweeps. However, it is not shown whether the subjects have hearing
problem. Maybe the events occurred because of the errors or artifacts during
recording.
6.5
Performance Analysis
As described in section 5.5, the sensitivity and specificity of CWT and IE
were calculated to test the performance of both techniques. By assuming that the
amplitudes in the ranges 0.1 to 1.99 µV are normal as mentioned in section 3.5.2 in
the last paragraph, the sensitivity and specificity of CWT and IE were calculated.
Table 6.5 shows the sensitivity and specificity of CWT technique and Table 6.6
shows the sensitivity and specificity of IE technique. In Table 6.5, it shows that the
highest sensitivity of CWT technique is 0.75 and the lowest sensitivity is 0.5. The
highest specificity of CW technique is 1 and the lowest is 0.5. Table 6.6,
the
highest sensitivity of IE technique is 1 and the lowest is 0.5. The highest specificity
of IE technique is 1 and the lowest is 0.42. After averaged all twelve values of
sensitivity and specificity, the IE technique has the sensitivity (0.88) higher than
CWT technique (0.63), but CWT technique has the specificity (0.83) higher than IE
technique (0.81). For overall performance, we can conclude that, IE technique
performed better than CWT technique.
81
Table 6.5: Sensitivity and specificity of CWT technique
Sweeps
Sensitivity
Specificity
1000
0.5
1
800
0.5
1
625
0.5
1
500
0.75
0.92
250
0.5
0.92
125
0.75
0.83
100
0.5
0.92
80
0.75
0.92
50
0.5
0.83
40
0.75
0.67
20
0.75
0.5
10
0.75
0.5
Average
0.63
0.83
Table 6.6: Sensitivity and specificity of IE technique
Sweeps
Sensitivity
Specificity
1000
1
1
800
1
1
625
1
1
500
1
0.92
250
0.75
0.92
125
0.75
1
100
1
0.83
80
1
0.83
50
0.75
0.75
40
0.75
0.58
20
0.5
0.42
10
1
0.42
Average
0.88
0.81
82
6.6
Summary
This chapter discussed about the results that had been produced in this study.
. The discussion consists of the explanation about the results of averaging technique
that had been used in this study, the results of CWT and IE techniques that had been
used as wave V detection and also the performance test of CWT and IE techniques.
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS
7.1
Conclusions
This thesis addressed the problem of current hearing screening programs
which are still constrained to the time consuming problem. This research focused on
releasing these constraints and provides a basis on developing the fast detection of
wave V.
Current hearing screening have to go through several stages by using several
types of hearing screening equipments on different stages. There are several
problems faced in order to implement hearing screening as discussed in the first
chapter of this thesis. The TEOAE, DPOAE and ABR have been used in the system.
It means the patients have to go through several tests before getting the results. The
target of this research is to reduce the stages by only using the ABR equipment.
In this research, the wave V marker had been introduced as the most
prominent and robust wave that has been used as indicator of hearing loss. By
detecting the wave V in the first stage, it is hope that the hearing screening will
84
become more effective compared with the current system. In this research, the gTec
equipments were used to acquire the ABR signals, produced the trigger signals and
attenuate the stimulus. The data acquisition system used in this research has
maximum sampling rate of 38.4 kHz and 24 resolution bit. The sampling rate used in
the research was only 19.2 kHz and not the maximum. Matlab software has been
used to acquire and analyze the signals.
A study on the effect of the noise removal on various sweeps averaged
signals was carried out. The averaging technique had been used in this study since it
is the most popular method to extract the ABR wave signal from the recorded brain
signal. The 2000 epochs are commonly used in this technique. In this research, the
2000 epochs were successfully reduced. The power spectrum was plotted to see the
effect on the frequency contents. Results show that the power spectrum only shows
the difference of noise contents between the raw signal and the averaged signal and it
does not depend to the quantity of sweeps.
Investigation on the correlation between the averaged signal and the power
spectrum was also carried out. Results show that the power spectrum can be used to
explain the morphology of the ABR signals. A study on the wave V detection using
ABR signals was carried out. Three ABR waves with different quantities of epochs
were plotted in order to mark the wave V. The experiments on CWT and IE
techniques as the wave marker were also carried out. Results show that those
techniques can be used to mark the wave. The results of those techniques were
discussed in chapter 6.
85
7.2
Recommendations
In this section, some problems of the techniques used in this research were
identified. Some improvements and extension from the current work are suggested.
The first problem deals with the raw signals. Although the use of averaging
technique is able to increase the signal to noise ratio of the averaged signal, the noisy
signal still affects the morphology of the averaged signals. It is recommended that
the noisy signal be eliminated before the averaging process. By averaging the good
signal and removing the bad signal, it will improve the quality of the averaged signal.
In this research, there is lack of ABR database. More recording of the ABR
raw signals have to be done from time to time. It is suggested that the database
recordings use the higher sampling rate as far as possible. The higher sampling data
can be down sample, but the lower sampling data cannot to be up sample since it will
not produced good signal. It is important for Malaysia to have its own ABR database
which is dynamic and grows from time to time. Moreover, a Malaysian ABR
database can be an asset for researchers to go deeper into developing their own ABR
system.
The detection of wave V in this research was done manually. It is better if an
automatic detection is developed. Development of the automatic detection can reduce
the time consumption that had been spent to test the performance of the techniques in
this study. The large ABR database also helps to improve the performance test of the
techniques. The real time detection of the wave V is recommended instead of the
offline method as used in this research. The real time method hopefully can provide a
convenient system to the end user.
86
On the recording process, the wired system was used in this research. The
development of the wireless system will make the recording process more
convenient. It can also reduce the artifacts such as conducted electric,
electromagnetic-field induced and radio frequency noises. Conducted noises come
from AC power lines (50 Hz and its numerous harmonics) through the power cord, as
well as from computer and through the interface cable. Electric and magnetic fields
exist in any clinical environment. Especially strong electromagnetic fields are found
in non-shielded rooms: operating rooms (OR), intensive care units (ICU), neonatal
intensive care units (NICU), hospital wards, and doctors’ offices. They come from
surrounding electrical wiring and various equipments and they introduce electromagnetic interferences (EMI) in conventional ABR systems. EMI may be so strong
that they can make ABR testing very difficult or impossible. Electric fields come
from building wiring and electrical equipment, and generate voltage in leads. The
amount of contamination from the electric field increases with the length of the leads
and the inter-electrode impedance mismatch. Magnetic fields induce currents in the
three loops of lead wires: between non-inverting and ground, inverting and ground,
and non-inverting and inverting electrodes.
87
REFERENCES
American Speech-Language-Hearing Association (1987). Short latency auditory
evoked potentials. ASHA Practice Policy, 1987;
http://www.asha.org/docs/html/RP1987-00024.html
American Speech-Language-Hearing Association (1997). Guidelines for Audiologic
screening.; http ://www.asha.org/policy
Beattie, R.C., Beguwala, F.E., Mills, D.M. and Boyd, R.L. (1986). Latency and
Amplitude Effects of Electrode Placement On The Early Auditory Evoked
Response. Journal of Speech and Hearing Disorder, 51:63-70.
Bradley, A.P. and Wilson, W.J. (2004). Automated analysis of the auditory
brainstem response. . In Proceedings of the Intelligent Sensors, Sensor
Networks and Information Processing Conference, 541 – 545.
Brigham, E.O. and Morrow, R.E. (1967). The Fast Fourier Transform. IEEE
Spectrum, 4(12):63-70.
Bell, S.L., Smith, D.C., Allen, R., and Lutman, M.E. (2004). Recording the Middle
Latency Response of the Auditory Evoked Potential as a Measure of Depth of
Anaesthesia. A technical note. British Journal of Anaesthesia. 92 (3): 442-5.
Bell, R.D. and Manon-Espaillat, R. (1998). Electroencephalography and Evoked
Responses. In Jay H Stein (Ed.) Internal medicine. (5th ed.) (pp. 905-913) St.
Louis, Mosby: Elsevier Health Sciences.
88
Beltran,N. and Cornejo,J.M. (2003). Simultaneous transient evoked otoacoustic
emissions and auditory brainstem response for auditory evaluation. In
Proceedings of the 25th International Conference of the IEEE Engineering in
Medicine and Biology Society, 4:3222-3224
Boston, J.R. (1981). Spectra of Auditory Brainstem Responses and Spontaneous
EEG. IEEE Trans. on Biomedical Engineering, 28(4):334-341.
Burkard,R. (1994). The Use of Maximum Length Sequences to Obtain Brainstem
Auditory Evoked Responses at Rapid Rates of Stimulation. American Journal
of Audiology. 3: 16 - 20.
Burkard,R.F., Eggermont,J.J., Don, M.(2007). Auditory evoked potentials: basic
principles and clinical application. Philadelphia : Lippincott Williams &
Wilkins.
Carr, J.J. and Brown, J.M. (2001). Introduction to Biomedical Equipment
Technology. (4th ed.) New Jersey, USA: Prentice Hall.
Cherrid,N., Naït-Ali, A. and Siarry,P.(2005). Optimized brainstem auditory evoked
potentials estimation using simulated annealing. Journal of Clinical Monitoring
and Computing, 19(3):231-238.
Dawson, D.M. (1998). Demyelinating Diseases. In Jay H Stein (Ed.) Internal
medicine. (5th ed.) (pp. 1007-1011) St. Louis, Mosby: Elsevier Health
Sciences.
Delb, W., D'Amelio, R., Schonecke, O. and Iro, H. (1999). Are there Psychological
or Audiological Parameters Determining the Tinnitus Impact. In J. W. P.
Hazell, editor, Proceedings of the 6th tinnitus seminar, Cambridge, UK,
Oxford University Press 446-451.
Delb, W. (2002). Universal Neonatal Hearing Screening in Germany. What is the
Next Step? HNO, 50:607-610.
89
Delb, W. (2003). Universal Neonatal Screening as an Application of Automated
Audiological Techniques. HNO, 51:962-965.
Delb, W., Merkel, D., Pilorget, K., Schmitt, J. and Plinkert, P. K. (2004).
Effectiveness of a TEOAE-based Screening Program. Can a Patient-Tracking
System Effectively be Organized using Modern Information Technology and
Central Data Management? Eur. Arch. Otorhinolaryngol., 261:191-196.
Delgado, R.E. and Ozdamar, O. (1994). Automated auditory brainstem response
interpretation. Engineering in Medicine and Biology Magazine, IEEE.
13(2):227-237.
Deuschl, G. (1999). Recommendations for the practice of clinical neurophysiology:
guidelines of the International Federation of Clinical Neurophysiology (2nd rev
and enlarged ed). Amsterdam: Elsevier.
Don, M.,and Kwong, B.(2002) Auditory Brainstem Response: Differential
Diagnosis. In: Jack Katz (Ed.) Handbook of Clinical Audiology. (5th ed.).(pp.
274-297) Philadelphia etc.: Lippincott Willams & Wilkins.
Drongelen, W.V. (2007) Signal processing for neuroscientists: introduction to the
analysis of physiological signals. Boston: Academic Press.
Enderle, J.D., Bronzino, J.D., and Blanchard, S.M. (2005). Introduction to
Biomedical Engineering.(2nd ed.). USA:Elsevier.
Gazella, G.S., Saini, S., Hahn, P.F., Goldberg, M.A. and Halpern, E.F. (1994). MR
Imaging of the Liver at 1.5T: value of signal averaging in suppressing motion
artifacts. American Journal of Roentgenology, 163: 335-337.
Hall, J.W. III. (1992). Handbook of Auditory Evoked Responses. Boston et al.: Allyn
and Bacon.
90
Habrakan, J.B.A, van Gils, M.J, and Cluitmans, P.J.M.(1993). Identification of Peak
V in Brainstem Auditory Evoked Potentials with Neural Networks. Computers
in Biology and Medicine, 23 (5):369-380.
Helfand, M., Thompson, D.C., Davis, R., McPhillips, H., Homer, C.J. and Lieu, T.L.
(2001). Newborn Hearing Screening, Systematic Evidence Review Number 5.
AHRQ Publication No 02-S001, Rockville MD Agency for Healthcare
Research and Quality.
Hood, L.( 1998). Clinical Applications of the Auditory Brainstem Response. San
Diego, London: Singular Publishing Group.
Kaplan, R. and Ozdamar, O. (1987), Microprocessor-based auditory brainstem
response (ABR) simulator. Medical and Biological Engineering and
Computing, 25(5): 560-566.
Kong,W., Donga, X. and Chen, Z. (2008). Voltage sag source location based on
instantaneous energy detection. Electric Power Systems Research,
78(11):1889-1898
Lakshmi Ananth (2009). “Infant Hearing Screening : Failing Hearing Tests Handling
Failed Hearing Test Results in Neonatal Hearing Screening” http://infanttoddlerhealth.suite101.com/article.cfm/infant_hearing_screening_failing_hearing_test
s Retrieved 2009-10-28
Lee, R. and Ozdamar, O. (1999). Analysis of wavelet preprocessed auditory
brainstem responses with self-organizing feature maps. Proceedings of The
First Joint BMES/EMBS Conference, 1:448.
Leski, J.M. (2002). Robust Weighted Avearging. IEEE Trans. on Biomedical
Engineering, 49:796--804.
91
Li, C., Zheng, C. and Tai, C. (1995). Detection of ECG Characteristic
Points Using Wavelet Transforms. IEEE Trans. on Biomedical Engineering,
42(1):21--28.
Licznerski,T.J., Kasprzak, H.T., and Kowalik, W. (1998). Analysis of Shearing
Interferograms of Tear Film Using Fast Fourier Transforms, J. Biomed. Opt. 3:
32.
Lightbody, G., McCullagh, P. J., and McAllister, H. G. (2006). Feature Extraction in
Auditory Brainstem Responses using Wavelet Decomposition on a Moving
Window of Waveform Data. Irish Signals and Systems Conference, 28-30
June, 2006. 321-326.
Loutridis, S.J. (2006). Instantaneous energy density as a feature for
gear fault detection. Mechanical Systems and Signal Processing, 20:1239–1253.
Maglione, J.L., Pincilotti, M., Acevedo, R.C., Bonell, C.E. and Gentiletti, G.G.
(2003). Estimation of the auditory brainstem response's wave V by means of
wavelet transform. Proceedings of the 25th International Conference of the
IEEE Engineering in Medicine and Biology Society, 3:2631-2634.
Marieb,E. N. (2001). Human Anatomy & Physiology. (5th ed.) San Francisco:
Addison Wesley Longman.
Malarvili,M.B, Kamarulafizam,I., Hussain, S. and Helmi,D. (2003). Heart sound
segmentation algorithm based on instantaneous energy of electrocardiogram.
Comput. Cardiol., 30:327-330.
Mason, S.M. and Adams, W. (1984). An Automated Microcomputer based Electric
Response Audiometry System for Machine Scoring of Auditory Potentials.
Clin. Phys. Physiol. Meas., 5:219-222.
McCarthy, B. (2007). Sound systems: design and optimization : modern techniques
and tools for sound system design and alignment. US: Focal.
92
McWilliams, A. (2008). “Hearing Loss - 4 Types of Hearing Tests to Detect Early”
http://ezinearticles.com/?Hearing-Loss---4-Types-of-Hearing-Tests-to-Detect-Early&id=1818571 Retrieved 2009-10-05
Meier, S., Narabayashi, O., Probst, R. and Schmuzinger, N. (2004). Comparison of
Currently Available Devices Designed for Newborn Hearing Screening using
Automated Auditory Brainstem and/or Otoacoustic Measurements. Int. J.
Pediatr. Otorhinolaryngol., 68:927-937.
Misulis,K.E. and Head, T.C.(2003). Essentials of clinical neurophysiology. (3rd ed)
UK: Bios Scientific.
Mohan, R. and C. Chui (1987). Use of Fast Fourier Transforms in Calculating Dose
Distributions for Irregularly Shaped Fields for Three-dimensional Traetment
Planning. Med.Phys., 14(1):70-77.
Morita, K. (1995). Applied Fourier Transform. Tokyo: Ohmsha Ltd.
Noback, C.R., Strominger,N.L., Demarest, R.J. (1991). The Human Nervous System.
(4th ed.) Philadelphia: Lea & Febiger.
Ontario Association For Families Of Children With Communication Disorders
(2008). "Speech and Language Terms and Abbreviations".
http://www.oafccd.com/factshee/fact59.htm. Retrieved 2008-12-02
Peters, J. G. (1986). The ALGO--1: An Automated Infant Hearing Screener Utilizing
Advanced Evoked Response Technology. Hear J., 89:335-353.
Peters, T.M., Williams, J.C. and Bates, J.H.T (1998). The Fourier Transform in
Biomedical Engineering. Boston: Birkhauser.
93
Plinkert, P. K. and Delb, W. (2001). Electronic Data Processing-Assisted
Organization of Interdisciplinary Universal Hearing Screening in Saarland.
HNO, 49:888-894.
Shangkai, G. and Loew, M. H. (1986). An Autoregressive Model of the BAEP Signal
for Hearing Threshold Testing. IEEE Trans. on Biomedical Engineering,
33:560--565.
Sininger, Y.,and Cone-Wesson, B.(2002) Threshold prediction using Auditory
Brainstem Response and Steady-State Evoked Potentials. In: Jack Katz (Ed)
Handbook of Clinical Audiology. (5th ed.).(pp. 298-322) Philadelphia etc :
Lippincott Willams & Wilkins.
Sokolov,Y., Kurtz,I., Steinman,A., Long,G., Sokolova,O.(2005).Vivosonic
IntegrityTM Technology: Making ABR Practical (Document 11115, Rev. 01).
Toronto, Canada: Vivosonic.
Sanei, S. and Chambers, J.A. (2007). EEG signal processing. Chichester, England;
Hoboken, NJ : John Wiley & Sons
Satoshi, I., Yasuhiro, H., Atsurou, S., Mituyoshi, N., Yasuyuki, H., Goro, O. and
Tomoyuki, H. (2003). A model of two-stage newborn hearing screening with
automated auditory brainstem response. Int. J. Pediatr. Otorhinolaryngol.,
67(10): 1099-1104.
Sharif, Z.; Zainal, M.S.; Sha`ameri, A.Z.; Salleh, S.H.S.; (2000) “Analysis and
classification of heart sounds and murmurs based on the instantaneous energy
and frequency estimations” TENCON 2000. Proceedings, 2000 Volume: 2
Page(s): 130-134.
Sh-Hussain, S. (2008). Different Techniques and Algorithms for Biomedical Signal
Processing. Malaysian Journal of Mathematical Sciences, 2(2): 133-145.
94
Sivalal, S. (2005). Screening for Hearing Loss in Infants. Health Technology
Assessment Report 2005, Malaysian Ministry of Health.
Sohmer,H., Kinarti, R. and Gafni, M. (1981). The latency of auditory nervebrainstem responses in sensorineural hearing loss. European Archives of OtoRhino-Laryngology, 230(2): 189-199.
Strauss, D. J., Delb, W., Plinkert, P. K. and Schmidt, H. (2004). Fast Detection of
Wave V in ABRs using a Smart Single Sweep Analysis System. In
Proceedings of the 26th International Conference of the IEEE Engineering in
Medicine and Biology Society, 458--461.
Swanepoel,D.W., Hugo, R. and Roode, R. (2004). Auditory Steady-State Responses
for Children With Severe to Profound Hearing Loss. Arch Otolaryngol Head
Neck Surg., 130(5):531-535.
Tan, H.G.R. and Mok, V.H. (2008). A Comparative Evaluation on Performance of
Coifman Discrete and Stationary Wavelet Transform in ECG Signal Denoise
Application. In Osman, N.A.A. and Ibrahim, F. and Abas, W.A.B.W. and
Rahman, H.S.A. and Ting, H.N. IFMBE Proceedings, 1, Volume 21, 4th Kuala
Lumpur International Conference on Biomedical Engineering 2008: BIOMED
2008, 25-28 June 2008, Kuala Lumpur, Malaysia (pp. 98-102). Berlin:
Springer.
Tan, Z.B., Wang, L,Y. and Wang, H. (2009). Signal Averaging for Noise Reduction
in Anesthesia Monitoring and Control with Communication Channels. Journal
of Biomedical Science and Engineering,
Tatum,W.O., Husain,A.M., Benbadis, S.R. and Kaplan, P.W. (2008). Handbook of
EEG interpretation.New York: Demos Medical Pub.
Wicke, J. D., Goff, W. R., Wallace, J. D. and Allison, T. (1978). On--line Statistical
Detection of Average Evoked Potentials: Application to Evoked Response
Audiometry. Electroenceph. Clin. Neurophysiol., 44:328-343.
95
Wilson, W.J., Winter, M., Kert, G. and Aghdasi, F. (1998). Signal Processing of the
Auditory Brainstem Response: Investigating into the use of Discrete Wavelet
Analysis. Procs. South African IEEE Symposium on Communications and
Signal Processing - COMSIG'98, pp 17-22.
Wilson, W.J. and Aghdasi, F.(1999a). Discrete wavelet analysis of the auditory
brainstem response: Effects of subject age, gender and test ear. AFRICON,
IEEE, 1:291-296.
Wilson, W.J. and Aghdasi, F.(1999b). Fast Fourier transform analysis of the auditory
brainstem response: effects of stimulus intensity and subject age, gender and
test ear. AFRICON, IEEE, 1:285-290.
Woodworth, W., Reisman, S. and Fointaine, A. B. (1983). The Detection of Auditory
Evoked Responses using a Matched Filter. IEEE Trans. on Biomedical
Engineering, 20:369-376.
Yoganathan, A.P., Gupta, R., Udwalia, F.E., Miller, J.W., Corcoran, W.H., Sarma,
R., Johnson, J.L. and Bing, R.J. (1976). Use of fast Fourier transform in the
frequency analysis of the first heart sound in normal man, Med. Biol. Eng.
Comput. 14, pp. 69–73.
Yoshinaga-Itano, C. (1999). Benefits of Early Intervention for Children with Hearing
Loss. Otolaryngol. Clin. North Am., 32:1089-1102.
Yost,W.A. (2007). Fundamental of Hearing: an Introduction. (5th ed.) London, UK:
Elsevier.
Zhang, R., McAllister, G., Scotney, B., McClean, S. and Houston, G. (2005). A
Comparison of Wavelet and Wavelet-Bayesian Approaches for the
Classification of Auditory Brainstem Response. The 3rd IEE International
Seminar on Medical Applications of Signal Processing, 65-70.
96
Zhang, R., McAllister,G.,Scotney,B.,McClean,S. and Houston,G.(2004). Feature
Extraction and Classification of the Auditory Brainstem Response Using
Wavelet Analysis. In López, J.A., Benfenati, E. and.Dubitzky,W. (Eds.):
Knowledge Exploration in Life Sciences Informatics (KELSI) ( pp. 169-180).
Berlin: Springer.
97
APPENDIX A
gTec EQUIPMENT SPECIFICATIONS
certified medical device
3.0
USB 2.0 interface
0636
FDA
medical
device
clearance
k060803
g.USBamp works with any type of passive electrodes and strip
or grid ECoG electrodes. The g.GAMMAsys active electrode
system can be used as well. Single cell recordings can be
performed in combination with
g.tec's spike sensor system.
of the MathWorks Inc.
are registered trademarks
MATLAB® and SIMULINK®
(The MathWorks, Inc., Natick, MA.)
g.tec is an official MATLAB partner.
g.USBamp can be used with a medical power supply or with a
rechargable battery pack for up to 10 hours of
independent operation.
CE and FDA certified medical device for non-invasive and invasive recordings
various software solutions available (driver/API, recording software, MATLAB/SIMULINK/LabVIEW ...)
4 independent ground potentials per unit to avoid interference between different signal types
internal amplifier calibration and automatic electrode impedance check
16 input channels per unit, units can be stacked to set up multi-channel systems
internal floating point DSP for digital preprocessing and signal filtering
24-bit high resolution ADCs, up to 38.4 kHz sampling with simultaneous S&H for all channels
Multiple units of g.USBamp can be stacked to set up a multichannel system. All channels are sampled synchronously.
notified body
generation
real DC-coupled EEG/ECoG/ECG/EMG/EOG biosignal amplifier with wide-range inputs
USB BIOSIGNAL AMPLIFIER
USBamp
Highlights
®
EEG/ECoG/ECG/EMG/EOG/... high performance biosignal acquisition
DC
DC
calib.
unit
imped.
unit
block
input
DC
DC
isolation
DC
DC
RECHARGABLE
BATTERY
PACK
DAC
ADC
DC
DC
5 V DC
isolation
MEMORY
USB 2.0
USB
controller
USB BIOSIGNAL AMPLIFIER
control and
processing
unit,
DSP
MEDICAL
POWER
SUPPLY
~ 100 - 240 V 50/60 Hz
SC / BLOCK
DIG I/O
SYNC OUT
SYNC IN
USB 2.0
Sensitivity:
< 30 nV (LSB) - ± 250 mV
Standards:
Directive of
medical products:
Safety class:
Applied part:
Size:
EN60601-1: 1996
(+A1 +A2 +A12 +A13)
EN60601-2-26: 2004
EN60601-1-2: 2003
EN60601-2-25 +A1: 2001
EN60601-2-40: 1998
93/42/EWG
II
type CF
197 x 155 x 40 mm
1000 g
standard safety
connectors and
system connectors
Weight:
> 100 MOhm
Input connectors:
< 0.3 µV RMS (0.1 - 10 Hz)
16 mono- / 8 bi-polar
(per device, softwareselectable)
Input impedance:
Noise level:
Input channels:
24 bit (38.4 kHz internal
sampling per channel)
12 bit
2 x DAC:
real DC coupled
16 x ADC:
Amplifier type:
enabled
High-Speed Online Processing for SIMULINK (or LabVIEW):
Online/real-time biosignal processing and recording with the maximum
system speed! g.USBamp appears as a block usable in any SIMULINK model.
The design of the hardware-interrupt controlled driver allows immediate
starting of the model without prior compilation. Also g.tec's specialized
g.RTanalyze blockset can be used for real-time parameter extraction and data
classification. The example shows a BCI system (P300-spelling device) with
g.USBamp realized in a SIMULINK model.
For offline biosignal analysis please see g.BSanalyze. This
software package includes powerful toolboxes for EEG analysis,
high-resolution EEG, ECG (heart rate and HRV analysis) and
single beat ECG analysis as well as for biosignal classification.
g.Recorder:
Our recording software supports all data acquisition devices
provided by g.tec. Comfortable system configuration, data
visualization and storage make g.Recorder a perfect tool for
teaching, research and clinical investigation. g.Recorder also
supports video-EEG and online biosignal parameter
monitoring.
The MATLAB-API:
With the MATLAB API the MATLAB Data Acquisition Toolbox can be used to get full access to the recording buffer and to use the whole functionality
of g.USBamp. The Data Acquisition Toolbox enables a quick and easy implementation of data visualization, processing and storage applications
under MATLAB.
API /device driver:
This option enables the integration of the hardware into an existing data recording or processing system by the user or to program applications in
C++ or other Windows-based programming languages. g.USBamp is also supported by BCI 2000.
Software options:
g.USBamp is equipped with 8 TTL-trigger inputs which are sampled synchronously with all input channels. Also additional digital I/Os are
accessable via a rear-side socket. The SC (short cut) input allows to disconnect the electrode sockets from the amplifiers during electrical or
magnetical stimulation in order to reduce artifacts.
DAC OUT
CALIB. OUT
AMPLIFIER
INPUT
(ELECTRODES)
CH 1 - 16
BLOCK DIAGRAM
rear side view
USB BIOSIGNAL AMPLIFIER
USBamp
101
APPENDIX B
PUBLICATIONS
Adeela Arooj, M.M. Rushaidin, Sh-Hussain Salleh and M.Hafizi Omar (2010). "Use of
Instantaneous Energy of ABR Signals for Fast Detection of Wave V", Journal of
Biomedical Science and Engineering 3 (8): 816-821.
M.M. Rushaidin, Sh-Hussain Salleh, Tan Tian Swee, J.M. Najeb and Adeela Arooj
(2009). "Wave V Detection Using Instantaneous Energy of Auditory Brainstem
Response Signal", American Journal of Applied Sciences 6 (9): 1669-1674.
M.M, Rushaidin, Sheikh Hussain Shaikh Salleh, J.M, Najeb and Tan Tian Swee (2008).
“Hardware Design for Low Cost Auditory Brainstem Response (ABR) Machine”
Proc. International Graduate Conference on Engineering and Science IGCES 2008.
Johor : UTM. 23-24 December 2008.
Tan Tian Swee, Sheikh Hussain Shaikh Salleh, J.M, Najeb, M.M, Rushaidin, and A.S,
Aisyah (2008). “Automatic Computerized Audiometric System” Proc.
International Graduate Conference on Engineering and Science IGCES 2008.
Johor : UTM. 23-24 December 2008.