Note: Descriptions are shown in the official language in which they were submitted.
CA 02269992 1999-04-23
Docket 65748
METHOD AND APPARATUS FOR COUGH SOUND ANALYSIS
Field of the Invention
This invention relates to methods and apparatuses for the analysis of
patient's
coughs. More specifically, this invention relates to methods and apparatuses
for the
analysis of patient's coughs to aid in diagnosing pulmonary disorders and
diseases.
This method uses signal analysis techniques to extract quantitative
information from
recorded cough sound pressure waves. Moreover, the method allows the
recordation
of cough sound waves while avoiding distortions caused by reflections. The
generated
data can be used to diagnose pulmonary disorders and diseases as well as track
the
effectiveness of treatment regimens over time. The method can also be used for
screening the general population, or populations at higher risk, so that such
pulmonary
disorders and diseases can be detected as early as possible so that
appropriate
treatment can be started as soon as possible.
Background of the Invention
Cough is associated with well over 100 different pulmonary diseases and is one
of the most common signs or symptoms of respiratory disease. Even though cough
may be an unwanted complication of a pulmonary disease, it has often been used
by
physicians as an effective diagnostic tool. Since cough sounds are composed of
acoustic information which can be altered by lung disease and since cough has
essentially the same acoustical characteristics whether performed voluntarily
or
involuntarily, analysis of voluntary cough sounds has the potential to become
a useful
noninvasive tool for screening large populations of workers to evaluate their
pulmonary function. The use of cough sound analysis to aid in the
identification of
lung disease has several distinct advantages since testing can be quickly and
easily
administered while requiring only a minimum amount of technician or patient
training.
In order to describe the events that occur during a cough, physiologists have
subdivided a cough into 4 different phases (Leith et al., Cough, In: The
Handbook of
Physiology, The Respiratory System edited by A. Fishman, P. T. Macklem and J.
Mead, Bethesda, MD, Am. Physiological Society, Sec(3) Chapter 20, 315-336
-1-
CA 02269992 1999-04-23
Docket 65748
(1987)). During the initial phase, called the inspiration phase, a variable
volume of air
is inhaled into the lungs. The second phase, referred to as the compression
phase,
begins as the glottis closes and the muscles of expiration begin to contract
increasing
thoracic pressure. The third phase is called the expulsion phase. At the start
of the
third phase, the glottis opens and gas flows rapidly from the lung. During the
fourth
and final phase, called the cessation phase, muscle activity is reduced and
airflow is
diminished.
The physical characteristics of a cough are illustrated in Figure 1. Flow from
the mouth during a cough is shown in Figure 1 A. Positive values represent
flow from
the lungs while negative flow values indicate airflow into the lungs. During
the initial
phase of a cough (phase I) airflow is negative as air enters the lungs. The
volume of
air inspired is variable and is said to be a function of the anticipated
forcefulness of the
cough (Yanagihara et al., "The Physical Parameters of Cough: the Larynx in a
Normal
Single Cough," Acta Oto-laryngol. 61: 495-510 (1966)). As compression of air
occurs during phase II of the cough, the glottis closes and airflow ceases.
When the
glottis reopens, in approximately 200 ms, flow initially increases and then
decreases
rapidly creating a flow transient. This initial rapid change in flow during
phase III is
referred to as supramaximal flow and is thought to result from the air rapidly
leaving
the flexible airway system as the airways compress during the initial part of
the
expulsion phase of a cough. At the same time that air is leaving the airways
during the
initial portion of phase III, expiratory flow from the lung periphery rises
sharply to
maximal flow which is limited by the maximum expiratory flow-volume
relationship
that is unique for each lung. Airflow leaving the lungs during a cough,
therefore, is a
summation of the transient air leaving the airways at a supramaximal flow rate
and the
air leaving the periphery of the lungs at maximal flow. During the cessation
phase IV
of a cough, airflow from the lungs diminishes and then approaches zero as
muscle
activity decreases.
Figure 1B illustrates a typical sound pressure wave generated by a cough. It
has been suggested that the cough sounds are generated during phase III and
sometimes during phase IV of a cough. The cough sound, itself, can be
subdivided
into two and sometimes three parts (Thorpe et al., "Towards a Quantitative
-2-
CA 02269992 1999-04-23
Docket 65748
Description of Asthmatic Cough Sounds," Eur. Respir. J. 5: 685-692 (1992)).
The
first part of a cough sound is referred to as the initial burst and represents
the sound
transient that is associated with the glottis opening. The second or middle
part
corresponds to the interval of near steady, maximal flow coming from the
periphery of
the lung which occurs with the glottis maximally open. The third part of a
cough,
called the final burst, is not always present, but is believed to occur in
some subjects
who close their glottis during the cessation phase of a cough.
Airflow from the lung during a cough and the maximum expiratory flow
volume (MEFV) relationship of a lung have much in common. During a forced
expiration the airways, which are very flexible cylindrical structures,
undergo
compression, and decrease in cross-sectional area as air rapidly passes
through them.
As a result, one or more choke points are created in the airway system during
maximal
gas flow. After a choke point has formed, flow from the lungs becomes
independent
of the driving pressure. This is important because it implies that airflow
through the
airway system should become effort independent during the performance of a
MEFV
maneuver. Once effort independence is reached, the MEFV relationship becomes
repeatable. Flow-volume curves recorded during a MEFV maneuver define the
limits
of flow and volume that can be achieved during most expiratory maneuvers in a
given
individual. Leith et al. (1987) have stated that a surprisingly modest
expiratory effort
is required to reach the outer limits of the flow-volume domain for a given
individual,
making forced expiration a reliable pulmonary function test. Figure 2 shows an
example of an MEFV curve while expiring with a maximum effort into a
spirometer.
An example of the flow volume relationship of a lung during a cough is
superimposed
on the MEFV curve in Figure 3. During the initial phase of a cough, air is
inspired
into the lungs. This is indicated by the increase in lung volume as the
operating point
on the flow-volume curve moves to the left throughout phase I. During the
compression phase, there is no gas flow so phase II is represented by a single
point on
the diagram. During the initial part of phase III, a supramaximal flow
transient is
observed as the volume of air in the flexible airways decreases quickly as the
airways
begin to collapse. Following the very brief flow transient, maximal flow is
achieved
which approaches the maximal flow reached during the performance of a MEFV
-3-
CA 02269992 1999-04-23
Docket 65748
maneuver. The events that occur during this portion of the cough are very
similar to
those that occur during a MEFV maneuver; therefore, it can be assumed that
airflow
leaving the lung during a cough reaches flow limitation and is reasonably
reproducible
when successive coughs are performed beginning at the same lung volume. Since
the
mechanisms producing cough sounds are dependent upon airflow, it seems likely
that
cough sounds are also reproducible if similar lung histories are followed
prior to each
cough.
A block diagram of a simple model illustrating how cough sounds are produced
is shown in Figure 4. It is thought that the acceleration and turbulence of
air within
the airways caused by the rapid expulsion of air from the lungs generates band
limited
noise which is then modified by the resonances of the lungs' upper airways and
possibly the oral and nasal cavities as air travels toward the mouth. Peaks
that occur
in the spectra of cough sounds result from resonances along the cough sound
pathway,
and they are similar to the formants observed in speech analysis. A second
source of
sound is referred to as a wheeze and is thought to result from the fluttering
of airway
walls as gas moves rapidly through the airways. A similar model has previously
been
proposed for the study of unvoiced speech (Oppenheim et al., "The Speech
Model" in
Discrete-time Signal Processing, New Jersey, Prentice Hall, Chapter 12, 816-
825
(1989)).
In the past, several groups of investigators have recorded cough sounds in a
variety of ways and have attempted to develop a technique which could be used
to
show differences in cough sounds between healthy subjects and those with
respiratory
diseases. It was thought that any substantial differences between coughs could
eventually become useful in identifying persons with respiratory diseases.
These initial
studies examined the coughs of subjects with a variety of obstructive lung
diseases, but
the most often studied population was that having asthma.
Debreczeni et al. ("Spectral Analysis of Cough Sounds Recorded With and
Without a Nose Clip," Bull. Eur. Physiopathol. Respir. Suppl. No. 10, 57s-61 s
(1987); "Spectra of the Voluntary First Cough Sound," Acta Physiol. Hung.
75(2):
117-131 (1990)) recorded sound pressure waves and computed the average spectra
of
cough sounds from patients with several types of obstructive lung disease. The
-4-
CA 02269992 1999-04-23
Docket 65748
spectra were used to determine how cough sound energy was distributed within
several frequency bands and to distinguish between coughs from healthy
subjects and
those subjects with lung disease.
Piirila et al. ("Differences in Acoustic and Dynamic Characteristics of
Spontaneous Cough in Pulmonary Diseases," Chest 96: 46-53 (1989)) recorded
cough
sound pressure waves and airflow for a sequence of coughs and then computed
average cough spectra. The peak values representative of the dominant
frequency
components within the cough were extracted for comparison. Spectrograms were
also
computed and the length of the cough maneuver was measured by determining the
time that sound energy had a frequency component present at 500 Hz. They
reported
that cough sound duration was longer for asthmatic coughs than for coughs from
control subjects. One problem with interpreting these studies, however, is
that
sequences of coughs were studied instead of single coughs. As a result, it has
been
difficult to interpret the measurements and conclusions of this study
concerning the
duration of a cough.
Researchers have studied several aspects of recording and interpreting cough
sounds. Initially, waterfall plots of cough spectrograms were examined to
determine if
there were differences in cough sounds of children with asthma before and
after
exercise (Toop et al., "Cough Sound Analysis: A New Tool for the Diagnosis of
Asthma?", Family Pract., 6(2): 83-85 (1989)). More recently, attempts have
been
made to evaluate cough sounds using more quantitative methods so that
meaningful
comparisons could be made between cough sounds of selected individuals (Thorpe
et
al., "Towards a Quantitative Description of Asthmatic Cough Sounds," Eur.
Respir. J.
5: 685-692 (1992)). The cough sound was divided into two or three parts and
the
characteristics of each part was studied independently. Power spectra were
computed,
normalized, and treated as histograms. The mean frequency, standard deviation,
skewness, and kurtosis were calculated. In addition, the energy within
selected
frequency bands was examined and differences in coughs from persons with
several
types of obstructive lung disease were noted. This study also examined
features of the
sound pressure wave with respect to time including the duration, root mean
square
(RMS) value and zero crossing rate. Interestingly, no significant difference
in the total
-5-
CA 02269992 1999-04-23
. =
Docket 65748
duration of a cough between healthy subjects and those with obstructive lung
diseases
was reported. They did note, however, that both the duration of the initial
burst and
zero crossing rates of the cough waveform during each of the first two phases
were
smaller for asthmatic than for non-asthmatic coughs.
A variety of methods have been used to record and digitize cough sounds for
analysis. The methods that have been used, however, have a large influence on
the
quality of the signal used for the acoustical analysis. Debreczeii, et al.
(1987, 1990)
recorded cough sounds of a seated subject in a quiet room with a microphone
directed
towards the subject's mouth from a distance of 50 cm. Sounds were recorded on
an
analog tape recorder and then digitized at a rate of 5 kHz and 20 kHz. Piirila
et al.
(1989) recorded cough sounds from subjects in a sitting position with a
microphone
attached to the skin of their chest wall located over the sternal manubrium.
The
sounds were recorded with a tape recorder; the bandwidth of their spectral
analysis
was 9 kHz. Toop et al. ("A Portable System for the Spectral Analysis of Cough
Sounds in Asthma," J. ofAsthma 27(6) 393-397 (1990)) described the design of a
system used to record cough sounds. A patient coughed into a tube with a
pneumotach and microphone attached. The cough signal was digitized at 5 kHz
for
analysis with a personal computer which limited the bandwidth of their
analysis to
frequencies below 2.5 kHz. Since the pneumotach modified the characteristics
of the
sound pressure wave reaching the microphone, these investigators estimated the
correct acoustical response by deconvolving their average spectral
measurements by
Weiner filtering.
In spite of past efforts, it would be desirable to provide a simple, reliable,
and
fast diagnostic method to analyze coughs in order to assist physicians in
diagnosing
lung disorders or diseases. It is also desirable to provide a simple,
reliable, and fast
diagnostic method to analyze coughs which will allow physicians to monitor the
effectiveness of treatments prescribed for lung disorders or diseases. It is
also
desirable to provide a simple, reliable, and fast diagnostic method to analyze
coughs
which can be used to screen populations for lung disorders or diseases,
especially in
cases where such lung disorders or diseases can be detected in an early stage
where
treatments can be more effectively administered and damage to lung function
can be
-6-
CA 02269992 1999-04-23
. =
Docket 65748
avoided or minimized. The methods and apparatuses of this invention provide
such
diagnostic methods.
Summarv of the Invention
This invention provides a fast, simple, reliable method and apparatus for
recording cough sounds for diagnostic and other medical purposes. More
specifically,
this invention relates to methods and apparatuses for the analysis of patient
coughs to
aid in diagnosing pulmonary disorders and diseases. This method uses signal
analysis
techniques to extract quantitative information from recorded cough sound
pressure
waves. The generated data can be used to diagnose pulmonary disorders and
diseases
as well as track the effectiveness of treatment regimens over time. The method
can
also be used to quickly and reliably screen individuals at risk of pulmonary
disorders
and diseases. The discovery of early stages of pulmonary disorders or diseases
may
allow earlier treatment and/or environmental modification to reduce the risk
of
irreversible injury to pulmonary function.
The present invention provides a method for analyzing coughs for diagnostic
purposes. This invention also provides a system for recording high fidelity
cough
sound measurements. Moreover, this invention provides a simple, non-invasive
system
that can quickly and easily be administered with minimum technician and
patient
training. The system comprises a mouthpiece, a tube having a distal end and a
proximal end, a flexible tubing having a distal end and a proximal end, and a
microphone; wherein the mouthpiece is attached to the proximal end of the
tube,
wherein the distal end of the tube is attached to the proximal end of the
flexible tube,
wherein the microphone is attached to the tube between its distal and proximal
ends
such that the microphone can record sound pressure waves within the system
without
distorting the pressure waves, and wherein the flexible tubing is sufficiently
long so
there are essentially no reflected sound pressure waves which interfere with
the
recording of the sound pressure waves at the microphone. Preferably, the
system also
includes a computer system to assist in recording and analyzing the sound
pressure
waves. Preferably, the distal end of the flexible tubing has an anechoic
termination to
further reduce or attenuate reflected sound waves.
-7-
CA 02269992 1999-04-23
Docket 65748
This invention also provides a method for analyzing a patient's cough for
diagnostic purposes, said method comprises (1) providing a system for
analyzing
coughs wherein the system comprises a mouthpiece, a tube having a distal end
and a
proximal end, a flexible tubing having a distal end and a proximal end, and a
microphone; wherein the mouthpiece is attached to the proximal end of the
tube,
wherein the distal end of the tube is attached to the proximal end of the
flexible tube,
wherein the nvcrophone is attached to the tube between its distal and proximal
ends
such that the nvcrophone can record sound pressure waves within the system
without
distorting the pressure waves, and wherein the flexible tubing is sufficiently
long so
there are essentially no reflected sound pressure waves which interfere with
the
recording of the sound pressure waves at the microphone; (2) allowing the
patient to
cough into the mouthpiece; (3) recording the sound pressure waves generated by
the
patient's cough with the nucrophone; and (4) analyzing the recorded sound
pressure
waves. Preferably, the recorded sound pressure waves are digitized and then
analyzed.
Preferably, the recorded sound pressure waves are analyzed using spectrograms
from
which contour plots can be generated.
These and other objectives and advantages of the present invention will be
apparent to those of ordinary skill in the art upon consideration of the
present
specification.
Brief Description of the Drawings
Figure 1 illustrates the physical events during a typical cough (Panel A) and
the
sound pressure waves generated by such a typical cough (Panel B).
Figure 2 provides an example of maximum expiratory flow volume (MEFV)
curve with a patient expiring with maximum effort into a spirometer.
Figure 3 illustrates the relationship of MEFV (Figure 2) with the physical
events of a typical cough (Figure 1).
Figure 4 provides a simple block diagram illustrating how typical cough sounds
are produced.
Figure 5 illustrates two methods for measuring sound pressure waves. Panel A
illustrates the interference in the recorded sound waves from reflections from
the
-8-
CA 02269992 1999-04-23
Docket 65748
surrounding environment. Panel B illustrates the method of this invention
which
allows the interference to be essentially eliminated.
Figure 6 illustrates a typical cough sound pressure wave generated using the
apparatus and method of the present invention.
Figures 7 and 8 provide examples of spectrograms generated from the sound
pressure wave of Figure 6. Figure 7 provides the frequency content of the
cough on
the vertical axis versus time on the horizontal axis; amplitude or intensity
of individual
frequency components are plotted on a logarithmic scale using the color scale
on the
right hand side of the Figure. Figure 8 provides the same data in a waterfall
type plot
where the spectral intensity is plotted on a logarithmic scale.
Figure 9 illustrates a three-dimensional autocorrelation plot of the data from
Figure 7.
Figure 10 shows the contour plot of the transitional region of the
autocorrelation plot shown in Figure 9 and corresponds to levels that are 1,
2, 3, 4,
and 5 percent of the peak level. This contour plot illustrates the
autocorrelation of a
spectrogram obtained from a normal subject.
Figure 11 provides contour plots similar to Figure 10 from four patients
diagnosed with obstructive lung diseases. The plots numbered 1 to 4 represent
data
from different patients.
Figure 12 provides contour plots similar to Figure 10 from four patients
diagnosed with restrictive lung diseases. The plots numbered 1 to 4 represent
data
from different patients.
Figure 13 plots the cough sound or acoustic moment index as determined by
the present method for control patients (normal) and patients with impaired
lung
function.
Figure 14 illustrates a system of this invention for recording cough pressure
waves using two microphones.
Figure 15 illustrates the system of the present invention combined with a
whole
body plethysmograph.
Figure 16 illustrates the system of this invention showing the data
acquisition
system, computer, and an exponential horn for terminating the flexible tubing.
-9-
CA 02269992 1999-04-23
Docket 65748
Description of the Preferred Embodiments
This invention provides a method and a system for analyzing cough sounds for
diagnostic purposes. A diagram illustrating the present system to measure
cough
sounds is shown in Figure 5B. A subject being tested coughs into a mouthpiece
connected to a tube to which one or more microphones have been attached. The
microphone is positioned so that its diaphragm is tangent to the inner surface
of the
tube and in a manner which minimizes reflections and/or distortion of sound
waves
passing through the tube. A long flexible section of hose or tubing is
connected to the
end of the metal tube opposite the mouthpiece. The system is designed so that
the
acoustical signal representing a cough sound pressure wave travels along the
tube as a
plane wave. As the sound travels through the long tubular system having a
constant
cross-section, it becomes attenuated. Any sounds reflected from the open tube
back
toward the microphone will be reduced in amplitude, therefore, and should not
significantly interfere with cough sounds measured at the microphone. This can
be
compared with recording cough sounds in a room (Figure 5A) in which
reflections
from the walls, ceiling, floor, and objects in the room interfere with the
recorded
cough sounds. Moreover, further reflections are generated within the mouth as
the
sound waves passes from the mouth into the room with the system in Figure 5A;
since
the cross-sectional area dramatically increases (i.e., from the mouth to
room), these
reflections and/or distortions can be significant. The reflections from the
walls, other
objects, and within the mouth itself are very difficult to eliminate. The
present system
significantly reduces and/or essentially eliminates such reflections or
distortions. By
using the essentially uniform cross-section of the mouthpiece, rigid tubing,
and flexible
tubing, the reflections from wall or objects in the room are essentially
eliminated.
Moreover, since the mouthpiece and the mouth opening are essentially the same
diameter when the subject coughs, reflections from this junction are also
significantly
reduced and minimized (especially relative to the situation illustrated in
Figure 5A).
A sound pressure wave generated during a cough can be digitized and recorded
using a sound analyzer. The digitized signal can then be transferred to a
computer for
analysis. An example of a cough sound pressure wave recorded with this system
is
shown in Figure 6. The wave is complex and has many frequency components which
-10-
CA 02269992 1999-04-23
Docket 65748
change as a function of time. Since it is difficult to quantify differences
between
acoustical signals in the time domain by visual inspection only, spectrograms
of the
time signals have often been computed. Spectrograms show the frequency content
of
a signal versus time. An example showing the spectrogram of the sound pressure
wave of the cough shown in Figure 6 is shown in Figure 7. The frequency
content of
the cough is plotted on the vertical axis versus time on the horizontal axis.
The
amplitude, or intensity, of individual frequency components of the cough are
shown in
terms of a logarithmic scale (shown by color) given on the right side of the
Figure.
The same information is plotted differently in Figure 8 where the spectral
intensity of
the same cough is plotted on a logarithmic (DB) scale as a waterfall plot.
Even though
differences between coughs are more easily visualized with spectrograms, it
still
remains difficult to quantify the differences and likenesses between cough
sounds from
different sources.
A unique method to compare cough sounds would simultaneously compare
how acoustical energy is distributed within a cough with respect to both time
and
frequency. An index, which uses a two dimensional autocorrelation function,
describes how this information is distributed within a spectrogram image. The
result is
threee-dimensional autocorrelation image. An example of the normalized
autocorrelation function of the spectrogram in Figure 7 is shown in a mesh
style plot in
Figure 9. The autocorrelation function forms a mountainous type surface with a
peak
at zero displacement (time (ti) and frequency (w) are both zero). As ti and w
increase,
the threee-dimensional mountainous surface approaches a plane surface. The
mountain reflects regions in which the energy of the cough sound is located
while the
plains represent regions with little or no sound energy. The transitional
region
between the plain and the mountain is unique for each cough. The transitional
region
of the autocorrelation plot shown in Figure 9 is shown as a contour plot in
Figure 10.
This autocorrelation plot is for a normal or control cough. The contours shown
correspond to levels that are 1, 2, 3, 4, and 5% of the peak level. The
autocorrelation
plot of the spectrogram of this normal subject forms a generally symmetrical
surface in
the horizontal and vertical directions. This normal or control contour plot
can be
compared with contour plots of the autocorrelation function of coughs from
patients
-11-
CA 02269992 1999-04-23
Docket 65748
who are known to have obstructive lung disease (Figure 11) and restrictive
lung
disease (Figure 12).
The contour plots have different shapes for subjects having obstructive and
restrictive lung diseases when compared to a healthy subject. In the case of
obstructive lung disease, the plots are stretched out in the horizontal
direction while
remaining the same or decreasing in the vertical direction. In contrast, the
plots from
persons with restrictive lung disease are reduced in the horizontal direction
while
remaining the same or increasing in the vertical direction. Thus, the contour
plots
obtained from data produced by this invention provide a simplified method to
diagnose
lung disorders or diseases.
In order to further simplify the analysis and provide a numerical estimation
of
pulmonary function, a cough sound index (CSI) or acoustic moment index was
defined
that was based on the shape of the transition contours of the autocorrelation
of the
spectrogram. This index was calculated using a moment analysis. Moments have
often been used in the past to describe the shape characteristics of
probability
distribution functions (Kendel et al., The Advanced Theory of Statistics, Vol
1,
Distribution Theory, New York, Hafner (1952)), the center of mass and the
rotational
properties of solid objects (Sears et al., "Rotation" in University Physics,
Massachusetts, Addison-Wesley Publishing Company, Inc., Chapter 9, 157-180
(1956)) and even the analysis of the shape of flow volume curves of the lung
(Becklake et al., "Evaluation of Tests of Lung Function for `Screening' for
Early
Detection of Chronic Obstructive Lung Disease" in The Lung in the Transition
Between Health and Disease, edited by P.T. Macklem and S. Permutt, New York
and
Basel, Marcel Decker, Chapter 16, 113-152 (1979)). Several moments of the
autocorrelation of the spectrogram of a subject with obstructive lung disease
illustrated in Figure 11-1 have been calculated about the horizontal and
vertical axis.
The various equations of this moment analysis of the autocorrelation function,
as well
as specific values for the specific patient in Figure 11-1 are as follows:
-12-
CA 02269992 1999-04-23
Docket 65748
Equations Example (Figure 11-1)
0' Moment (Area)
M.0 = f f(x)dx M,,O = 0.0576
M~O = f f(y)dy Myo = 0.576
1 ` Moment (Mean)
M.1= f x-f(x)dx Mx1= 0.520
M,1=fY'f(Y)dY My1=0.514
2' Moment (Variance)
M,2 = f(x-x)Z-f(x)dx Mx2 = 0.00102
M,2 = f (Y-Y)2,f(y)dy MY2 = 0.0000818
3`d Moment (Skewness)
M,3 = f(x-x)3=gx)dx M~3 = 3.95 x 10''
M,3 = f(Y-Y)3.f(Y)dY K3 - 2.98 X 10
4' Moment (Kurtosis)
Mic4 = f(x-x)4=f(x)dx Mic4 = 2.28 x 10'5
7
K4 = f(Y-Y)4'f(Y)dY MY4 = 2.64 x 10'
The second moment is used in the definition of the cough sound index (CSI).
The
second moment My2 is equivalent to the moment inertia of the autocorrelation
about
the horizontal axis and M,,2 is equivalent to the moment of inertia of the
autocorrelation about the vertical axis. The CSI is calculated as follows:
(1) when M,,2 / Mn < 1, then
CSI = 10 - [(M.2 / My2)- 1]
and
(2) when M,,2 / My2 z 1, then
CSI=(M.2 /My2)-1
-13-
CA 02269992 1999-04-23
Docket 65748
The CSI index tends to be large and positive for persons with obstructive lung
disease
and negative for persons with restrictive lung disease. Normal subjects should
have a
cough index near zero. In the case of the patient from Figure 11-1, M.2 / My2
is 12.5
and, therefore, the CSI is 11.5, which indicated an obstructive lung disease
in
agreement with the clinical evaluation. The other moments indicated above
could be
used to further define the shape of the autocorrelation function and could
also be used
for diagnostic purposes.
Patients with obstructive and restrictive lung diseases who volunteered to
have
their cough sounds analyzed while being tested at the pulmonary clinic of the
Department of Pulmonary Medicine at West Virginia University School of
Medicine
were tested. Their cough sound or acoustic moment indices were computed and
compared witll their clinical diagnosis based on spirometry measurements. The
results
are shown in Figure 13. Although the statistical analysis has not been
completed, this
study shows good agreement between the clinical diagnosis and the diagnosis
made
using the cough sound index.
Figures 5B and 16 illustrate the present system using one microphone. A
subject being tested coughs into a mouthpiece 10 connected to a tube 12 to
which one
or more microphones 14 have been attached. Preferably the mouthpiece 10 is
adapted
to fit the mouth of the patient; normally the mouthpiece has an inside
diameter of
about one inch. Preferably the mouthpiece 10 is disposable so that a new one
can be
used for each patient. Of course, other sized mouthpieces can also be used;
for
example, the mouthpiece could be in the range of about 0.5 to about 2 inches
in
diameter. Indeed, for testing children, smaller diameter mouthpieces may be
preferred;
in such case, the other components should be modified to have the same
diameter and
cross-sectional area. The mouthpiece 10 is attached to a tube 12 which is
preferably
rigid. The tube 12 is preferably constructed out of rigid plastic or metal and
has a
circular cross-section without any tapenng. Normally, the inside diameter of
tube 12
is comparable to that of mouthpiece 10. Thus, the inside diameter of tube 12
is
preferably about one inch although it could vary from about 0.5 to about 2
inches in
diameter. The mouthpiece 10 is attached to the proximal end of the tube 12.
Generally the tube 12 is about 6 to 18 inches long; other lengths can be used
if desired.
-14-
CA 02269992 1999-04-23
Docket 65748
A long flexible tube 16 is attached to the distal end of the tube 12 using
coupling 20.
Preferably, coupling or joint 20, as well as other joints within the system,
are
essentially "seamless" so as to minimize reflections as the sound waves pass
the joints.
Located between the proximal and distal ends of the tube 12 is the microphone
14.
The microphone 14 is generally perpendicular to the tube such that its
diaphragm is
essentially tangential to, and essentially flush with, the inner surface of
the tube.
Preferably the microphone 14 is sufficiently small so that it does not
significantly
distort or interfere with the sound waves as they pass from the patient down
the tube.
A 1/4 inch microphone (Bruel & Kjaer, Mode14136) has been found acceptable. As
shown in Figure 14, more than one microphone can be used if desired. Such a
system
may further limit the effect of reflections and/or other distortions.
Generally, the length of the flexible tube is adjusted so that (1) reflections
of
sound waves back toward the microphone are significantly reduced or minimized
and
(2) the back pressure or resistance within the tubing is not sufficient to
significantly
distort the cough. Generally the length of the flexible tubing 16 is about 2
to about 50
feet. More preferably, the length is about 10 to about 25 feet; for a inner
diameter of
about one inch, a length of about 15 feet appears to give reliable and
reproducible
results. The flexible tubing can be looped or coiled so long as it is not
kinked or
otherwise significantly distorted. Suitable materials for the flexible tubing
include, for
example, gum rubber, neoprene, hypalon, silicone, santoprene, tygon, latex,
norprene,
and the like. The flexible tubing 16 preferably has the same inside diameter
and cross-
section (e.g., circular) as the tube 12. Preferably the joints or couplings
(e.g., coupling
20) between the various components are essentially "seamless" to avoid
distortions to
the sound wave as it passes through the system. The distal end of the flexible
tubing
16 is open. Preferably the distal end of the flexible tubing 16 is terminated
with an
anechoic termination (e.g., an exponential horn 18 as shown in Figure 16).
The main function of the flexible tubing is to attenuate the sound signal in
order to reduce reflections. After the "true" (i.e., reflection and distortion
free) signal
is expelled from the mouth, it travels through the mouthpiece and is recorded
by the
microphone in the rigid tube. At this point, it would be ideal to effectively
make the
sound "disappear" so as to eliminate any reflections of the signal that could
be picked
-15-
CA 02269992 1999-04-23
Docket 65748
up by the microphone. Such reflections would appear as noise in the recording
and
would be difficult to separate from the true signal. The flexible tubing
provides an
outlet for the sound to travel into and a means to attenuate the signal. Since
the
flexible tubing is approximately the same inside diameter as the rigid tube
(along with
the mouthpiece and mouth opening), the acoustic impedance mismatch is
minimized,
thereby significantly reducing reflections that might occur at the junction of
the rigid
tube and the flexible tube. As the sound travels down the flexible tube, part
of the
energy is absorbed by the tube, thereby attenuating the signal. The more the
sound is
attenuated the less signal remains which can be reflected. When the sound
reaches the
end of the tube, if open, there is a large cross-sectional area change (i.e.,
diameter of
the tubing to size of room) which can cause a considerable reflection. To
better match
cross-sectional areas, the distal end of the flexible tube is terminated in an
anechoic
termination to further attenuate reflections. Thus, for example, an
exponential horn 18
(e.g., trombone bell or horn) can be added at the distal end of the flexible
tube as
shown in Figure 16. The portion of the signal that is reflected then travels
back up the
flexible tube (where it is further attenuated); any reflected signal remaining
and which
reaches the microphone will be recorded as unwanted noise on top of the
signal.
Theoretically, if the tube was long enough, the reflected signal would not
"come back"
to the microphone until the recording period (i.e, normally one second) was
over. This
is not feasible, however, since substantially increasing the tubing length
increases
resistance to flow which can produce unnatural coughs. Thus, there is a
tradeoff in
noise reduction (i.e., reduction in reflected sound) and increased resistance.
With
inside diameters of about one inch throughout the system, a length of about 2
to about
50 feet is acceptable with a length of about 10 to about 25 feet being
preferred, and a
length of about 15 feet being most preferred.
Types of flexible tubing that are suitable for the present invention include,
for
example, gum rubber, neoprene, hypalon, silicone, santoprene, tygon, latex,
norprene,
and the like. The major factors affecting the different types of flexible
tubing appears
to be their sound absorbency and their "loading effects" (i.e., distortion of
the signal
due to the acoustical properties of the tubing itself). Thus, with each type
of tubing
there is a tradeoff between attenuation and loading. Latex tubing appears to
give the
-16-
CA 02269992 2007-11-27
greatest attenuation; tygon tubing appears to provide the least loading; and
neoprene
appears to provide the best combination of the two factors. The preferred
tubing at
the present time is, therefore, latex, tygon, and neoprene.
The mouthpiece 10, tube 12, microphone 14, flexible tube 16, and coupling 20
are designed to minimize the distortion of the pressure waves moving through
the
system. Thus, they should be as closely matched as possible with regard to the
inside
diameter and cross-sectional area to provide essentially seamless transitions.
As
discussed above, these components typically have an inside diameter of about
0.5 to 2
inches, preferably about 0.75 to 1.5 inches, and most preferably about 1 inch.
The
cross-section throughout the system should be uniform and preferably is
circular.
Thus, the acoustical signal representing a cough sound pressure wave travels
along the
various components as a plane wave. As the sound travels through the long
tubular
system having a constant cross-section, it becomes attenuated. Any sounds
reflected
from the open tube back toward the microphone will be reduced in amplitude,
therefore, and do not significantly interfere with cough sounds measured at
the
microphone. For comparison purposes, Figure 5A illustrates recording cough
sounds
in a room in which reflections from the walls, floor, ceiling, and objects in
the room
interfere with the recorded cough sounds. These reflections are very difficult
to
eliniinate. The present system significantly reduces, and essentially
eliminates, such
reflections or distortions. The use of the exponential horn as shown in Figure
16 can
be, and preferably is, used to reduce the reflections and distortions even
further.
Additional attempts were made to modify the shape of the termination of the
flexible
tubing in order to even further reduce these reflection and distortions. For
example, a
conical wedge was placed at different locations within horn 18 in Figure 16 to
determine if distortions could be further reduced; the conical wedge had
little or no
effect. As shown in Figure 15, the present system can also be used in
conjunction with
conventional plethysmograph techniques to measure lung volume. Thus, effects
of
lung volume on the cough sound index can be determined.
The following examples are intended to illustrate the invention and not to
limit it.
-17-
CA 02269992 1999-04-23
Docket 65748
Example 1. A cough analyzer as illustrated in Figure 5B was constructed using
a cylindrical mouthpiece attached to a one inch (i.d.) metal tube (eleven
inches long).
A 1/4 inch microphone (Bruel & Kjaer, Model 4136) was mounted at 90 on the
metal
tube with its diaphragm tangent (i.e., flush) with the inner surface of the
metal tube. A
15 foot section of flexible latex tubing was attached to the distal end of the
tube. The
flexible tubing was open at its distal end. Sound pressure waves from
voluntary
coughs were collected using the microphone and then digitized and recorded
using a
sound analyzer and computer. A five pole high pass Butterworth filter was
applied to
the data to reduce the effect of frequencies below about 50 Hz.
Cough sound measurements were obtained for 21 patients, including four
controls; seven patients with restrictive lung disease, and 10 patients with
obstructive
lung disease based on conventional clinical examinations. Cough durations were
defined as the time during which 0.05 to 99.95 percent of the cough energy
occurred.
The maximum energy frequency of the coughs was defined as the frequency where
the
maximum amount of energy occurred. The high frequency was the frequency below
which 99.95 percent of the energy occurred during the cough. A cough sound
index
(CSI) was determined from the shape of the autocorrelation of the joint time-
frequency
spectrogram of the cough. The following results were obtained:
Controls Restrictive Lung Obstructive Lung
(n=4) Disease (n=7) Disease n=10
Duration (msec) 391 f 44 369 f 79 544 51
Maximum 285 f 24 212 f 24 225 f 29
Energy
Frequency (Hz)
High Frequency 4428 f 363 4052 f 309 3821 f 235
(]Elz)
CSI -0.97 f 1.30 -1.37 f 1.82 5.34 t 1.97
Patterns were evident between the different disease types, especially when
considering
obstructive lung diseases.
-18-
CA 02269992 1999-04-23
Docket 65748
Example 2. Using equipment and procedures similar to Example 1, sound
pressure waves from 25 volunteers (four controls; nine with restrictive lung
disease;
ten with marked obstructive lung disease; and two with niild obstructive lung
disease)
were obtained. A moment analysis of the contour representing a 97 percent
decrease
in the maximum value of the autocorrelation function was performed along both
the
Ati and Oc,o axes. The calculated second moments were used to determine the
cough
sound index (CSI). The following results were obtained:
Subject CSI
Controls +0.08 0.6
Restriction -2.11 1.5
Mild Obstruction +0.31 0.6
Marked Obstruction +5.85 2.3
The results showed good agreement between cough sound analysis and clinical
findings.
-19-
