A Robust Multichannel Lung Sound Recording Device

Elmar Messner

, Martin Hagm¨uller

, Paul Swatek

and Franz Pernkopf

Signal Processing and Speech Communication Laboratory, Graz University of Technology, Graz, Austria

Division of Thoracic and Hyperbaric Surgery, Medical University of Graz, Graz, Austria

Keywords:

Lung Sounds, Lung Sound Transducer, Air-Coupled Microphone, Multichannel Recording Device, Sound

Classiﬁcation.

Abstract:

This paper presents a robust multichannel lung sound recording device (LSRD) for automatic lung sound

classiﬁcation. Compared to common approaches, we improved the usability and the robustness against body

sounds and ambient noise. We developed a novel lung sound transducer (LST) and an appropriate attachment

method realized as a foam pad. For analogue preﬁltering, preampliﬁcation, and digitization of the lung sound

signal, we use a composition of low-cost standard audio recording equipment. Furthermore, we developed a

suitable recording software. In our experiments, we show the robustness of our LSRD against ambient noise,

and we demonstrate the achieved signal quality. The LST’s microphone features a signal-to-noise ratio of

SNR = 80 dB. Therefore, we obtain a bandwidth of up to a frequency of f ≈ 2500 Hz for vesicular lung

sound recordings. Compared to the attachment of the LST with self-adhesive tape, the foam pad achieves

an attenuation of ambient noise of up to 50 dB in the relevant frequency range. The result of this work is

a multichannel recording device, which enables a fast gathering of valuable lung sounds in noisy clinical

environments without impeding the daily routines.

1 INTRODUCTION

Computer-aided lung sound analysis offers advan-

tages for medical diagnosis, such as digital stor-

age, monitoring in critical care settings, computer-

supported analysis, and comparison among different

sound recordings. Despite these advantages, it is far

away from being used in clinical settings. One rea-

son is the lack of efﬁciency and performance due

to the variability in the recordings (Reichert et al.,

2008; Gurung et al., 2011). Recently, lung sound

research mainly focused on the classiﬁcation task.

Researchers either performed lung sound recordings

independently or used appropriate databases in their

experiments (Palaniappan et al., 2013). Sensors ap-

plied to lung sound recording are air-coupled mi-

crophones, contact sensors and modiﬁed stethoscope

chest pieces (Pasterkamp et al., 1993; Kraman et al.,

2006). The most common recording technique em-

ploys air-coupled microphones attached with self-

adhesive tape. This approach lacks of sensitivity

against body sounds and ambient noise (Zanartu et al.,

2009; Pasterkamp et al., 1999; Liu et al., 2013).

Moreover, for multichannel usage the attachment

of several lung sound transducers (LSTs) with self-

adhesive tape results in a poor usability and increases

the sensitivity against measurement errors.

In this paper, we introduce a robust lung sound

recording device (LSRD) to circumvent the afore-

mentioned drawbacks. It supports a reliable record-

ing of a high quality lung sound database for mul-

tichannel lung sound classiﬁcation. To obtain clean

lung sound recordings, we focused on the recording

stage and reduced post-processing. Furthermore, it

was important that the LSRD is suitable to record

lung sounds for a large number of diseases. This

is reﬂected in the distribution and position of the

LSTs. Beside distinct adventitious lung sounds (So-

vijarvi et al., 2000), it should reliably allow the esti-

mate of changes in amplitude of lung sounds, which

is necessary for the detection of, e.g., pneumotho-

rax (Hayashi, 2011).

Based on the approach with air-coupled micro-

phones (Pasterkamp et al., 1993), we developed a

novel lung sound transducer ensuring a high signal

quality. For the attachment of the LSTs, we designed

a foam pad similar to the Stethographics STG 16

(Murphy, 2007). It records lung sounds in supine

position. We implemented the analog preﬁltering,

preampliﬁcation, and digitization of the lung sound

signal with a composition of standard audio record-

ing equipment. The entire LSRD consists of the foam

Messner, E., Hagmüller, M., Swatek, P. and Pernkopf, F.

A Robust Multichannel Lung Sound Recording Device.

DOI: 10.5220/0005660200340039

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 1: BIODEVICES, pages 34-39

ISBN: 978-989-758-170-0

pad (our so called auscultation pad) and a pneumota-

chograph, both working with an appropriaterecording

software on a personal computer.

We organized the paper as follows. Section 2

presents our LST design and the auscultation pad.

Section 3 gives an overview on the remaining compo-

nents of the LSRD. The achieved signal quality and

the robustness against ambient noise is treated in Sec-

tion 4. Section 5 describes the measurement proce-

dure and Section 6 concludes the paper.

2 AUSCULTATION PAD

The core of our LSRD is the auscultation pad. It is a

foam pad with 16 LSTs distributed on its surface. We

adapted our LST design for this attachment method.

In the following subsections, we separately describe

the LST design and the foam pad.

2.1 Lung Sound Transducer

We developed a novel LST according to the ap-

proach with air-coupled electret-condenser micro-

phones (Pasterkamp et al., 1993). Figure 1 shows our

ﬁnal LST. We use a Littmann Classic II chest piece

as coupler. We inserted an electret-condenser micro-

phone capsule (ECMC) in such a way that the stetho-

scope chest piece is acting as a conical coupler be-

tween the microphone capsule and the human skin.

The depth of the conical coupler is d = 2.2 mm, and

the width is w = 33 mm. Its shape corresponds to

the recommendations in (Wodicka et al., 1994; Kra-

man et al., 1995). We used the diaphragm of the

chest piece to cover its opening. It prevents the ﬁll-

ing of the coupler cavity with skin, and, thus, it en-

sures its acoustic effect. This is important for varying

contact pressure, and, therefore, it is relevant for our

attachment method discussed in Section 2.2. The di-

aphragm further enables a convenient disinfection of

the LST, and it protects the ECMC from scratching

Sealing Ring

Diaphragm

Venting

Chest Piece

ECMC

Figure 1: Lung sound transducer consisting of an electret-

condenser microphone capsule (ECMC) inserted in a

Littmann Classic II chest piece.

body hair and dirt.

To allow static pressure equalization between the

coupler chamber and the surrounding air, we inserted

a small vent into the chest piece. We used a thin-wall

23-g needle with a length of l = 11.5 mm and with an

inner diameter of d = 0.35 mm, according to the rec-

ommendations in (Kraman et al., 1995). As ECMC,

we used the Primo EM172, which features a very high

signal-to-noise ratio of SNR = 80 dB and a sensitivity

of −27 dB (re 1V/Pa). These speciﬁcations are dis-

tinctly better than those of widely used microphones,

like the Sony ECM-44BPT (Sen and Kahya, 2006) or

the Sony ECM-77B (Dokur, 2009), which feature an

SNR ≤ 64 dB.

2.2 Pad

The attachment of the LST is a crucial part, because

of its high sensitivity against air- and tissue-borne

sounds (Zanartu et al., 2009; Pasterkamp et al., 1999).

Therefore, we developed a foam pad, the auscultation

pad shown in Figure 2. It consists of several foam lay-

6 cm

6 cm6 cm

7 cm

1 2

5 6

11 12

15 16

Figure 2: Arrangement of the lung sound transducers on the

auscultation pad. The center line represents the spine.

ers and a cover of artiﬁcial leather. The topmost lay-

ers holds the LSTs. There is a small cavity beneath

each LST, which avoids increasing contact pressure

due to the underlying foam. Furthermore, the cavity

prevents the foam from touching the venting of the

LST or even clogging it. By using different kinds of

foam, we designed a shape that adapts to almost ev-

ery physique. This construction provides a symmet-

ric contact pressure with respect to the spine. We ar-

A Robust Multichannel Lung Sound Recording Device

ranged the LSTs on the surface of the auscultation pad

with a ﬁxed pattern, which almost matches the one

proposed in (Sen and Kahya, 2006). The pad enables

a fast attachment of the LST on the posterior chest by

simply placing the auscultation pad under the back of

the patient in supine position. To further stabilize the

patient, we extendedthe auscultation pad with two ad-

ditional pads, one for the head and another one for the

buttocks, as shown in Figure 3. We are able to achieve

a high robustness against body sounds and ambient

noise with an overall high lung sound signal quality.

Further details are presented in Section 4. The attenu-

ation of ambient noise is due to the surrounding foam.

We achieve the robustness against body sounds due to

the reliable attachment and almost no movement of

the back during breathing in the supine position. The

surrounding foam further prevents the interspersal of

body-borne noise in the LST cable. Another advan-

tage is the balanced audio connection from the aus-

cultation pad to the microphone preampliﬁers, which

reduces the susceptibility to external noise.

3 REMAINING SETUP

COMPONENTS

We use the auscultation pad as part of a mobile

recording setup (Figure 3). The setup consists of an

equipment cart, which includes the recording hard-

ware, two screens, a pneumotachograph,and the pads.

The following subsections contain some details about

the remaining components.

Main Screen

Patient Screen

Auscultation Pad

Additional

Pads

Pneumotachograph

Figure 3: Mobile lung sound recording device containing

the auscultation pad and the remaining components.

3.1 Recording Hardware

We implemented the analogue preﬁltering, preampli-

ﬁcation, and digitization of the lung sound signal with

low-cost standard audio recording equipment. The

composition of the appropriate hardware fulﬁlls the

requirements of the CORSA guidelines (Vannuccini

et al., 2000).

We use two SM Pro Audio EP84 8-channel mi-

crophone preampliﬁers with the integrated ADAT in-

terface SM Pro Audio PR8IIA. Beside the high-pass

ﬁltering (cutoff frequency f

= 80 Hz, with a slope

of 18 dB/oct), the preampliﬁcation, and the analog-

to-digital conversion of the LST signal, the SM Pro

Audio PR8IIA provides the supply voltage (phantom

power) for the ECMCs. For the appropriate operat-

ing voltage of the ECMCs, we further use AKG MPA

VL phantom power adapters. They convert the phan-

tom power of 48V to the required 3V~10V. The AKG

MPA VL phantom power adapter features a high-pass

characteristic with a cutoff frequency of f

= 80 Hz

and a slope of 6 dB/oct. In a series-connection with

the microphone preampliﬁer, an overall high-pass

characteristic with a slope of 24 dB/oct is achieved.

The high-pass ﬁlters of the microphone preampliﬁers

and the phantom power adapters are implemented as

Bessel ﬁlters. Therefore, they feature an approxi-

mately linear phase response. The two SM Pro Au-

dio EP84 are connected with an RME Fireface 800

audio interface. This represents a ﬁrewire multichan-

nel audio recording device for a computer. For the

lung sound recordings we use a sampling frequency

of f

= 16 kHz and a resolution of 24 Bit.

3.2 Pneumotachograph

The simultaneous recording of the velocity of respired

air and the lung sounds makes the distinction between

inspiratory and expiratory phases possible. Further-

more, we almost reach a uniform lung sound signal

intensity proﬁle by specifying the respiratory behav-

ior of the patient, resulting in a high quality database.

We use a Schiller SP 260 pneumotachograph con-

nected via the USB port.

3.3 Recording Software

We developed a MATLAB GUI for the simultaneous

recording of the air ﬂow and the lung sounds. Beside

the recording task, it enables the gathering of meta-

data and the clinical report.

We use Playrec for the multichannel recording of

the lung sound signals with MATLAB. For the ﬂow

signal, we read the serial port of the pneumotacho-

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

Figure 4: MATLAB GUI for the recording of the air ﬂow

and the lung sounds signals.

graph. Figure 4 shows the main screen of the soft-

ware, featuring simultaneously recorded air ﬂow and

lung sounds. Furthermore, we have a patient screen

to display the measured air ﬂow in real time.

4 ROBUSTNESS AND SIGNAL

QUALITY

In this section, we show the achieved signal quality

by means of the SNR and the frequency range of the

recording setup. We further demonstrate the robust-

ness against ambient noise with a simple experiment.

4.1 Robustness Against Ambient Noise

The CORSA guidelines (Rossi et al., 2000) recom-

mend to have environmental condition with a back-

ground noise level of preferably < 45 dB(A) during

lung sound recordings. These requirements are not

always given in clinical setting.

We compared the performance of our auscultation

pad in a noisy setting with the attachment-method of

the LST with self-adhesive tape (Pasterkamp et al.,

1993). The experiment took place in a small room.

We considered two measurement scenarios. In the

ﬁrst scenario we centered the auscultation pad on the

ﬂoor with a test person lying on it. In the second sce-

nario, we placed a chair in the center of the room, with

the test person in a sitting position.

With self-adhesive tape, we attach an LST on the

persons posterior chest, at the same position where it

was attached with the auscultation pad. The resid-

ual signal acquisition chain remained the same for

both scenarios, as introduced in Section 3. As noise

−120

−110

−100

−90

−80

−70

−60

−50

Frequency [Hz]

PSD [dB]

Self-adhesive tape

Auscultation pad

Figure 5: Attenuation of background noise of the ausculta-

tion pad compared to the attachment method of lung sound

transducers with self adhesive tape.

sources, we used ﬁve loudspeakers, which played

back white Gaussian noise. The loudspeakers fea-

ture a ﬂat frequency responce from f = 80 Hz to

f = 20 kHz. We measured the A-weighted equiva-

lent sound level for both scenarios at the position of

the sensor over 30 seconds with L

Aeq

= 68 dB. During

the recording of the LST signal, we instructed the test

person to hold its breath for 15 seconds and played

back the white Gaussian noise.

Figure 5 shows the power spectral density (PSD)

for the LST signal in both scenarios in the relevant

frequency range. We see a distinct attenuation com-

pared to the attachment with self-adhesive tape, start-

ing above a frequency of f = 500 Hz. At a frequency

of f ≈ 2 kHz the difference is up to 50 dB.

4.2 Signal Quality

In Figure 6 we show the SNR of our recording setup

by illustrating the spectral characteristics of the vesic-

ular lung sound of a healthy adult person. The blue

line shows the spectral characteristics during the in-

spiratory phase. The red line shows the background

noise recorded at breath hold; the frequency compo-

nents in the low-frequency range are mainly caused

by body sounds.

0 500 1000 1500 2000 2500 3000

−120

−110

−100

−90

−80

−70

−60

−50

−40

Frequency [Hz]

PSD [dB]

Inspiratory lung sound

Background noise

Figure 6: Spectral characteristics of normal breath sounds

and background noise at breath hold, recorded at the poste-

rior chest of a healthy adult.

A Robust Multichannel Lung Sound Recording Device

We achieve a signal-to-noise ratio of up to SNR =

75 dB with an additional head room of around 5 dB.

Due to the high SNR value, we reach a bandwidth up

to a frequency of f ≈ 2500 Hz.

5 MEASUREMENT PROCEDURE

The lung sounds are recorded in the supine position

on an examination table. The auscultation pad is

placed under the back of the patient. For the orien-

tation of the patient on the pad, we use a deﬁned dis-

tance between the the 7th cervical (C7) vertebra and

the topmost row of sensors. The patient is instructed

to lie quietly on the table and to breath with a max-

imum inspiratory ﬂow of 1.5 l/s. This corresponds

to values used from the authors in (Jones et al., 1999;

Malmberg et al., 1995) and also the recommendations

in the CORSA standard (Vannuccini et al., 2000). The

recording time can be speciﬁed in the recording soft-

ware.

6 CONCLUSIONS

We developed a robust lung sound recording device

(LSRD), which reliably records a high quality lung

sound database for multichannel lung sound classiﬁ-

cation. With preliminary measurements, we success-

fully underline the robustness of our newly designed

auscultation pad with respect to ambient noise. Com-

pared to the attachment of the LST with self-adhesive

tape, we achieve an attenuation of ambient noise of

up to 50 dB in the relevant frequency range. Due

to the high signal-to-noise ratio of our LST’s micro-

phone of SNR = 80 dB, we obtain a bandwidth of up

to f = 2500 Hz for vesicular lung sounds. For care-

fully performed measurements, our LSRD reduces

the post-processing to band-pass ﬁltering and heart

sound reduction.

As future work, we plan to record a lung sound

database for several lung diseases. Furthermore, we

will focus on the classiﬁcation of lung sounds.

ACKNOWLEDGEMENTS

This project was supported by the govern-

ment of Styria, Austria, under the project call

HTI:Tech

for Med. The authors acknowledge 3M

for providing Littmann® stethoscope chest pieces

and Schiller AG for the support with a spirometry

solution.

REFERENCES

Dokur, Z. (2009). Respiratory sound classiﬁcation by us-

ing an incremental supervised neural network. Pattern

Analysis and Applications, 12(4):309–319.

Gurung, A., Scrafford, C. G., Tielsch, J. M., Levine, O. S.,

and Checkley, W. (2011). Computerized lung sound

analysis as diagnostic aid for the detection of ab-

normal lung sounds: a systematic review and meta-

analysis. Respiratory medicine, 105(9):1396–1403.

Hayashi, N. (2011). Detection of pneumothorax visualized

by computer analysis of bilateral respiratory sounds.

Yonago acta medica, 54(4):75.

Jones, A., Jones, R. D., Kwong, K., and Burns, Y. (1999).

Effect of positioning on recorded lung sound intensi-

ties in subjects without pulmonary dysfunction. Phys-

ical Therapy, 79(7):682–690.

Kraman, S. S., Wodicka, G. R., Oh, Y., and Pasterkamp,

H. (1995). Measurement of respiratory acoustic sig-

nals. Effect of microphone air cavity width, shape, and

venting. CHEST Journal, 108(4):1004–1008.

Kraman, S. S., Wodicka, G. R., Pressler, G. A., and

Pasterkamp, H. (2006). Comparison of lung sound

transducers using a bioacoustic transducer testing sys-

tem. Journal of Applied Physiology, 101(2):469–476.

Liu, S., Gao, R. X., John, D., Staudenmayer, J., and Freed-

son, P. (2013). Tissue artifact removal from respira-

tory signals based on empirical mode decomposition.

Annals of biomedical engineering, 41(5):1003–1015.

Malmberg, L. P., Pesu, L., and Sovij¨arvi, A. (1995). Sig-

niﬁcant differences in ﬂow standardised breath sound

spectra in patients with chronic obstructive pulmonary

disease, stable asthma, and healthy lungs. Thorax,

50(12):1285–1291.

Murphy, R. (2007). Development of acoustic instruments

for diagnosis and management of medical condi-

tions. Engineering in Medicine and Biology Maga-

zine, IEEE, 26(1):16–19.

Palaniappan, R., Sundaraj, K., and Ahamed, N. U. (2013).

Machine learning in lung sound analysis: a systematic

review. Biocybernetics and Biomedical Engineering,

33(3):129–135.

Pasterkamp, H., Kraman, S., DeFrain, P., and Wodicka,

G. (1993). Measurement of respiratory acoustical

signals. Comparison of sensors. CHEST Journal,

104(5):1518–1525.

Pasterkamp, H., Wodicka, G., and Kraman, S. (1999). Ef-

fect of ambient respiratory noise on the measurement

of lung sounds. Medical & Biological Engineering &

Computing, 37(4):461–465.

Reichert, S., Gass, R., Brandt, C., and Andr`es, E. (2008).

Analysis of respiratory sounds: state of the art. Clini-

cal medicine. Circulatory, respiratory and pulmonary

medicine, 2:45.

Rossi, M., Sovijarvi, A., Piirila, P., Vannuccini, L., Dal-

masso, F., and Vanderschoot, J. (2000). Environmen-

tal and subject conditions and breathing manoeuvres

for respiratory sound recordings. European Respira-

tory Review, 10(77):611–615.

BIODEVICES 2016 - 9th International Conference on Biomedical Electronics and Devices

Sen, I. and Kahya, Y. (2006). A multi-channel device for

respiratory sound data acquisition and transient detec-

tion. In Proceedings of the 27th Annual International

Conference of the IEEE Engineering in Medicine and

Biology Society (EMBS’06), pages 6658–6661.

Sovijarvi, A., Malmberg, L., Charbonneau, G., Vander-

schoot, J., Dalmasso, F., Sacco, C., Rossi, M., and

Earis, J. (2000). Characteristics of breath sounds and

adventitious respiratory sounds. European Respira-

tory Review, 10(77):591–596.

Vannuccini, L., Earis, J., Helisto, P., Cheetham, B., Rossi,

M., Sovijarvi, A., and Vanderschoot, J. (2000). Cap-

turing and preprocessing of respiratory sounds. Euro-

pean Respiratory Review, 10(77):616–620.

Wodicka, G. R., Kraman, S. S., Zenk, G. M., and

Pasterkamp, H. (1994). Measurement of respira-

tory acoustic signals. Effect of microphone air cavity

depth. CHEST Journal, 106(4):1140–1144.

Zanartu, M., Ho, J. C., Kraman, S. S., Pasterkamp, H., Hu-

ber, J. E., and Wodicka, G. R. (2009). Air-borne and

tissue-borne sensitivities of bioacoustic sensors used

on the skin surface. IEEE Transactions on Biomedi-

cal Engineering, 56(2):443–451.

A Robust Multichannel Lung Sound Recording Device