of lung sounds. We suggest that lung sound spectra should be averaged at known airfiows over several breaths .... recording of lung sounds after giving informed consent. The study protocol was approved ..... explanation for their rela- tively.
Measurement of respiratory acoustical signals. Comparison of sensors. H Pasterkamp, S S Kraman, P D DeFrain and G R Wodicka Chest 1993;104;1518-1525 The online version of this article, along with updated information and services can be found online on the World Wide Web at: http://chestjournal.chestpubs.org/content/104/5/1518
Chest is the official journal of the American College of Chest Physicians. It has been published monthly since 1935. Copyright1993by the American College of Chest Physicians, 3300 Dundee Road, Northbrook, IL 60062. All rights reserved. No part of this article or PDF may be reproduced or distributed without the prior written permission of the copyright holder. (http://chestjournal.chestpubs.org/site/misc/reprints.xhtml) ISSN:0012-3692
Downloaded from chestjournal.chestpubs.org by guest on July 14, 2011 © 1993 American College of Chest Physicians
Measurement Signals* Comparison
of Respiratory
Acoustical
of Sensors
,
M. D.; Steve S. Kraman, M.D. F C. C. F; M.S.E.E.; and George R. Wodicka, Ph.D.
Hans Pasterkamp, Paul D. DeFrain,
We assessed the performance of three air-coupled and four contact sensors under standardized conditions oflung sound recording. Recordings were obtained from three of the investigators at the best site on the posterior lower chest as determined by auscultation. Lung sounds were band-pass filtered between 100 and 2,000 Hz and sampled simultaneously with calibrated airflow at a rate of 10 kHz. Fourier techniques were used for power spectral analysis. Average spectra for inspiratory sounds at flows of 2 ± 0.5 Ifs were referenced against background noise at zero flow. Aircoupled and contact sensors had comparable maximum signal-to-noise ratios and gave similar values for most coustical
signals
from
are
assessed
ditionally
During
recent
number
of publications
tion
analysis
and
years,
Ofl
techniques
auditory
have
increasing
page
shown
proc-
1320
the
limits
frequency sounds
content
and
indicate
a potential
During sounds
bronchial are said
details
distribution value
provocation to show
application is in
where
a close
airway
patency also
upper
respiratory
and
correlation has
observed
Advances
for
central between
been
described
an
Department
meas-
airway
obstruction
sound
spectra
in physical
of
Pediatrics,
University
of
analysis
Unfortunately,
use
various
in-
methods
that
in
this
area
recently
has
been
.
and other in any recording
Advantages
and
sound sensors of respiratory
disadvantages
are fundaacoustical
of air-coupled-
versus contact-type sensors previously have been described but not formally compared in actual recording of respiratory sounds.6 We, therefore, decided to evaluate
the
selection techniques
ofsensors in situ. Also, the effect for processing of the recorded
sounds
relative
was
methods
performance
assessed,
for
and
of a representative
suggestions
presentation
of the
of different respiratory
of informative
acoustical
data
were
forth.
and
accelManitoba,
Canada (Dr. Pasterkamp); the VA Medical Center, Ky (Dr. Kraman); and the School of Electrical EngiPurdue University, West Lafayette, Ind (Mr. DeFrain and
Dr. Wodicka). This study was supported in part by a grant from the Whitaker Foundation and a National Science Foundation Young Investigator Award BCS-9257488 to Dr. Wodicka. Dr. Pasterkamp is supported by the Children’s Hospital ofWinnipeg Research Foundation. This study was presented in part at the 17th International Conference on Lung Sounds, August, 1992, Helsinki, Finland. Manuscript received December 3, 1992; revision accepted February 26, 1993. Reprint requests: Dr. Pasterkamp, CN503-840 Sherbrook Street, Winnipeg, Manitoba, Canada R3A 151
1518
the world
reaches
sound
laboratories.
standardization
SUBJECTS
models3 have
signal
differ from the choice of sound sensors, through the sampling and processing of sound signals, to the measurement and presentation of results. The need
put
clinically. technology
the lung sound
of respiratory
clinical
around
We used
in microprocessor
the
acoustical
introduction
into
emphasized Microphones mental parts
of human
at which
frequency noise level.
the
signals.
the
application. inspiratory
urements
erated
for
to provide
surpass
highest
=
background
vestigators
acquisisignal
F
systems
in median frequency that correlates with the of airflow obstruction even in the absence of both over the lung’ and at the trachea.2
Another
Winnipeg, Lexington, neering,
an
Digital
see
been that
respiratory
for clinical for example,
*Fmm
been
signals.
comment
sounds
Ofl
of normal
and
has
tra-
perception.
Studies
increase degree wheeze,
system auscultation.
on computer-assisted
of these
respiratory
respiratory
by subjective
there
For editorial essing
the
spectral parameters. Unexpectedly, less sensitivity (lower signal-to-noise ratio) at high frequencies was observed in the air-coupled devices. Sensor performance needs to be characterized in studies of lung sounds. We suggest that lung sound spectra should be averaged at known airfiows over several breaths and that all measurements should be reported relative to sounds recorded at zero flow. (Chest 1993; 104:1518-25)
used
seven
sensors
for respiration
acoustic
subjects
for
the
consent.
The study
Committee were
on
healthy
in height
from subjects
the study.
they
flow
range
intensity before
oflung
sounds
after All
commonly
of us served giving
by the Purdue
subjects.
was
sounds
the
flow
posterior
the experiment
three
Us.
as
informed
University participants
The chest
point was
Downloaded from chestjournal.chestpubs.org by guest on July 14, 2011 © 1993 American College of Chest Physicians
month
of Electrical
pneumotachograph The
of maximum identified
for successive
Acoustical
the
sat in a soundproof
on an oscilloscope.
marked
of Respiratory
during
at the School
subjects
a calibrated
lower and
infection
place
The signal
at 2 ± 0.5
Measurement
took
through the
set
tract
University.
breathed
observed over
ofthose
1). Three
was approved
of human
a respiratory
Purdue and
while
lion
had
recording
chamber
of lung
protocol use
(Table
male nonsmokers, ranging in age from 24 to 47 years, 166 to 183 cm, and in weight from 62 to 83 kg. None
before The
METHODS
studies
recording the
of the
Engineering,
AND
that are representative
target sound
by auscultaplacement
Signals(Pastes*amp
of
eta!)
Table
ofSound
1-Dimensions
Sensors
before
spectral
calculated Height, Sensors
Diameter,
g
overlap
mm
the
155*
ECM77*
12.0
1.7
5.6
12.0
1.5
5.6
8.9
2.0
7.6
RadioShackNo.33-1052t HP 210501
EMT25C
PPG
No.
power
52.2
14.0
15.4
28.0
Digital
8.0
9.9
28.0
5.1
2.1
20.0
FYSPac2II *sony
Corp.
Montvale,
,
tRadio Shack, Hewlett-Packard, §Siemens,
Tandy
Iselin,
IIPPG
lung
of lung
sounds
sound
(below
which
band
Tex.
and
below
than
NJ.
(phonopneumography)
sensor,
University,
Technion
Belgium.
sensors
were
first placed
were
with
affixed
in coupling
double-sided
diameter,
pressure
20 mm.
by
placement
directly
while
chambers.
to
recording
chamber
bore
hole.
a closed-cell with
of the
taped
The
a lateral
microphone was mm in diameter,
The
the
chest was
cavity The
a central
was
coupler
foam
disk,
hole
air-coupled
plastic
where
this
sound
spectra
Lung
Radio
and
On
31
(range, contained number
in diameter
for
sensor
was
surface with standard masking tape. kept the same for each subject throughout
The the
amplified
four-pole
Butterworth-type
digitized
at 10 kHz
converter
(4801a,
ware
used
was
Calif).
provided
acquisition,
personal
computer
control
analysis
of the
sound
fast signals.
flow
using
signals
were
Customized and
(NB3,
signal Hz,
analog-to-digital
Mass).
playback
respiratory
2,000
a 12-bit
analysis,
and ensured
2,048-point
and
Woburn,
data
least four complete used
with
Digital-to-analog
quality
We
Corp.,
100 and
Sound
channel
ADAC
to the individual
between filters.
per
for
IBM-compatible Torrance,
according
band-pass-filtered
of the
artifact-free
soft-
display
Epson
on
America
an Inc.,
transforms
A Hanning
data
sound
rec)rded
sounds
recording
over
for power window
Table
at
spectral
was
applied
2-Perftwmance
Subject
above
We
to less
calculated
the
and the frequency
and
band
analysis
attenuation
significance
of the
dropped
regression
plot
to the of power
(decibel/octave).
were
assessed
by re-
Newman-Keuls was
multiple
accepted
when
prob-
0.05.
average, 16.0
the to 30.3
length
six inspirations of spectra within 18 to
spectra The
from an average slopes of the
sounds
recorded compared
sensors
(Table
trates
these
with
ofSound
1
the background
those
with
findings noise signal
20.8
sensor,
s and
(range, 4 to 9). The average the target flow range was 34
recorded
2). A significantly observed
background lung sound
was and
noise
of3l samples (range, 9 to 70). spectral curves of inspiratory air-coupled microphones were
with
steeper
was
subject
57). We computed
(range,
uation
of recording
s) for each
in
with
greater the
contact
sound
Sony
electret
attenuation (p2,000
25.0
-
17.5
955
>2,000
26.0
-
12.4
765
Sensors
signal
sensors
of variance
Statistical
than
limits
frequencies
noise.
linear
phones (p2,000 935
ratio.
plot from 300 to 700 Hz. reaches
background
noise
level.
CHEST
I
104
I 5 I NOVEMBER,
Downloaded from chestjournal.chestpubs.org by guest on July 14, 2011 © 1993 American College of Chest Physicians
1993
1519
0.01
Subject
No. 3
0.001
0.0001
I E-5
N
I LC)
I E-6
I E-7
I I E-8
I E-9
IE-lO 0
500
1000
1500
Frequency FIGURE 1. Comparison in situ of two signals and noise F1,, lowest frequency second quartile frequency (median greatest.
the
1,000 were
spectral curves Hz indicate, close
to
or
and missing data points therefore, that the lung at
background
mean
I 000
-
noise
three contact sensors, plotting difference between the lung sound signal reaches background noise level; F,,.,,, frequency at which the ratio of signal to noise is
above sounds
level.
(Hz)
and
air.cotipled at which frequency);
Q2,
2000
difference
500 and
The
this
case
in spectral 1,000
Hz.
was
obtained
slopes The
is most
greatest with
evident
between
signal
band
width
PPG
sensor.
For
the
in the
± SD
_______________________
V
1st quartile
U
median
0
3rd quartile
A
spectral
(Q1)
(Q2)
(Q3)
edge (SEw)
N I >
.1
0
C
a)
#{149}1
#{149}
C. U.
I
.
:
FIGURE
of2±0.5
1520
I
I
I
Sony
Sony
RadiO
ECM155
ECM77
#33-1052
2. Measurements Us.
on average
lung
Shack
sound
spectra,
FYSPaC2
obtained
with
HP
PPG
Siemens
21050
#201
EMT25C
seven
Measurement
sensors
at inspiratory
of Respiratory
Acoustical
Downloaded from chestjournal.chestpubs.org by guest on July 14, 2011 © 1993 American College of Chest Physicians
flows
Signals
(Pasterkemp
etaQ
30
25
-0-
EMT25C
(#2)
-,---
EMT25C
(#9)
-s-----
PPG
#201
20 #{149}0
15 U) 0
C 10
C C)
Cl)
No. 4
Subject
I 000
I 500
Frequency 3.
FIGURE
sensitivity EMT 25C).
FYSPac2 depicts
sensor, the
the
mechanical
Comparison (maximum
in situ of three contact signal-to-noise ratio) between
spectral
peak
resonance
near
of this
2,000
Hz
device.
sensors, two
(Hz) showing a significant difference in effective sensors of the same make and model (Siemens
significant used
a characteristic
The values
FYSPac2 transducer gave significantly higher for all spectral parameters compared with those
either
of the higher
other with
tation. Figure
sensors the Sony
(Fig and
2). The first Radio Shack
reduction
sensor
alternatively
quartile was microphones
of
sensitivity
in
for the measurements of
some
this
damage
4 illustrates
Siemens
2, reflecting
particular
sustained
our
the
in Table
sensor
during
observations
or
transpor-
in one
subject
compared with the HP, PPG, and Siemens sensors (p